{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.0, "global_step": 27732, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 9.012256669069935e-08, "loss": 0.8072, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.802451333813987e-07, "loss": 0.8046, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.7036770007209804e-07, "loss": 0.8047, "step": 15 }, { "epoch": 0.0, "learning_rate": 3.604902667627974e-07, "loss": 0.8054, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.5061283345349677e-07, "loss": 0.8012, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.407354001441961e-07, "loss": 0.7992, "step": 30 }, { "epoch": 0.01, "learning_rate": 6.308579668348954e-07, "loss": 0.7948, "step": 35 }, { "epoch": 0.01, "learning_rate": 7.209805335255948e-07, "loss": 0.7923, "step": 40 }, { "epoch": 0.01, "learning_rate": 8.111031002162942e-07, "loss": 0.7845, "step": 45 }, { "epoch": 0.01, "learning_rate": 9.012256669069935e-07, "loss": 0.7823, "step": 50 }, { "epoch": 0.01, "learning_rate": 9.913482335976928e-07, "loss": 0.769, "step": 55 }, { "epoch": 0.01, "learning_rate": 1.0814708002883922e-06, "loss": 0.7664, "step": 60 }, { "epoch": 0.01, "learning_rate": 1.1715933669790915e-06, "loss": 0.7475, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.261715933669791e-06, "loss": 0.732, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.3518385003604903e-06, "loss": 0.7131, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.4419610670511896e-06, "loss": 0.6831, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.532083633741889e-06, "loss": 0.6519, "step": 85 }, { "epoch": 0.01, "learning_rate": 1.6222062004325884e-06, "loss": 0.6234, "step": 90 }, { "epoch": 0.01, "learning_rate": 1.7123287671232877e-06, "loss": 0.5967, "step": 95 }, { "epoch": 0.01, "learning_rate": 1.802451333813987e-06, "loss": 0.5721, "step": 100 }, { "epoch": 0.02, "learning_rate": 1.8925739005046864e-06, "loss": 0.5463, "step": 105 }, { "epoch": 0.02, "learning_rate": 1.9826964671953856e-06, "loss": 0.5216, "step": 110 }, { "epoch": 0.02, "learning_rate": 2.072819033886085e-06, "loss": 0.5, "step": 115 }, { "epoch": 0.02, "learning_rate": 2.1629416005767843e-06, "loss": 0.4781, "step": 120 }, { "epoch": 0.02, "learning_rate": 2.253064167267484e-06, "loss": 0.4573, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.343186733958183e-06, "loss": 0.437, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.4333093006488826e-06, "loss": 0.4174, "step": 135 }, { "epoch": 0.02, "learning_rate": 2.523431867339582e-06, "loss": 0.399, "step": 140 }, { "epoch": 0.02, "learning_rate": 2.6135544340302814e-06, "loss": 0.3815, "step": 145 }, { "epoch": 0.02, "learning_rate": 2.7036770007209805e-06, "loss": 0.3626, "step": 150 }, { "epoch": 0.02, "learning_rate": 2.7937995674116797e-06, "loss": 0.3462, "step": 155 }, { "epoch": 0.02, "learning_rate": 2.8839221341023792e-06, "loss": 0.3297, "step": 160 }, { "epoch": 0.02, "learning_rate": 2.9740447007930784e-06, "loss": 0.3135, "step": 165 }, { "epoch": 0.02, "learning_rate": 3.064167267483778e-06, "loss": 0.2986, "step": 170 }, { "epoch": 0.03, "learning_rate": 3.154289834174477e-06, "loss": 0.2842, "step": 175 }, { "epoch": 0.03, "learning_rate": 3.2444124008651767e-06, "loss": 0.2708, "step": 180 }, { "epoch": 0.03, "learning_rate": 3.334534967555876e-06, "loss": 0.2575, "step": 185 }, { "epoch": 0.03, "learning_rate": 3.4246575342465754e-06, "loss": 0.245, "step": 190 }, { "epoch": 0.03, "learning_rate": 3.5147801009372746e-06, "loss": 0.2329, "step": 195 }, { "epoch": 0.03, "learning_rate": 3.604902667627974e-06, "loss": 0.2217, "step": 200 }, { "epoch": 0.03, "learning_rate": 3.6950252343186733e-06, "loss": 0.2105, "step": 205 }, { "epoch": 0.03, "learning_rate": 3.785147801009373e-06, "loss": 0.2006, "step": 210 }, { "epoch": 0.03, "learning_rate": 3.8752703677000725e-06, "loss": 0.1907, "step": 215 }, { "epoch": 0.03, "learning_rate": 3.965392934390771e-06, "loss": 0.1816, "step": 220 }, { "epoch": 0.03, "learning_rate": 4.055515501081471e-06, "loss": 0.1729, "step": 225 }, { "epoch": 0.03, "learning_rate": 4.14563806777217e-06, "loss": 0.1648, "step": 230 }, { "epoch": 0.03, "learning_rate": 4.235760634462869e-06, "loss": 0.1572, "step": 235 }, { "epoch": 0.03, "learning_rate": 4.325883201153569e-06, "loss": 0.1493, "step": 240 }, { "epoch": 0.04, "learning_rate": 4.416005767844268e-06, "loss": 0.1422, "step": 245 }, { "epoch": 0.04, "learning_rate": 4.506128334534968e-06, "loss": 0.1354, "step": 250 }, { "epoch": 0.04, "learning_rate": 4.5962509012256665e-06, "loss": 0.1294, "step": 255 }, { "epoch": 0.04, "learning_rate": 4.686373467916366e-06, "loss": 0.1235, "step": 260 }, { "epoch": 0.04, "learning_rate": 4.776496034607066e-06, "loss": 0.1177, "step": 265 }, { "epoch": 0.04, "learning_rate": 4.866618601297765e-06, "loss": 0.1122, "step": 270 }, { "epoch": 0.04, "learning_rate": 4.956741167988464e-06, "loss": 0.1073, "step": 275 }, { "epoch": 0.04, "learning_rate": 5.046863734679164e-06, "loss": 0.1023, "step": 280 }, { "epoch": 0.04, "learning_rate": 5.136986301369863e-06, "loss": 0.0983, "step": 285 }, { "epoch": 0.04, "learning_rate": 5.227108868060563e-06, "loss": 0.0934, "step": 290 }, { "epoch": 0.04, "learning_rate": 5.3172314347512615e-06, "loss": 0.0898, "step": 295 }, { "epoch": 0.04, "learning_rate": 5.407354001441961e-06, "loss": 0.0861, "step": 300 }, { "epoch": 0.04, "learning_rate": 5.497476568132661e-06, "loss": 0.0821, "step": 305 }, { "epoch": 0.04, "learning_rate": 5.587599134823359e-06, "loss": 0.0789, "step": 310 }, { "epoch": 0.05, "learning_rate": 5.677721701514059e-06, "loss": 0.0759, "step": 315 }, { "epoch": 0.05, "learning_rate": 5.7678442682047585e-06, "loss": 0.0726, "step": 320 }, { "epoch": 0.05, "learning_rate": 5.857966834895458e-06, "loss": 0.0697, "step": 325 }, { "epoch": 0.05, "learning_rate": 5.948089401586157e-06, "loss": 0.0667, "step": 330 }, { "epoch": 0.05, "learning_rate": 6.038211968276856e-06, "loss": 0.0638, "step": 335 }, { "epoch": 0.05, "learning_rate": 6.128334534967556e-06, "loss": 0.0616, "step": 340 }, { "epoch": 0.05, "learning_rate": 6.2184571016582555e-06, "loss": 0.0591, "step": 345 }, { "epoch": 0.05, "learning_rate": 6.308579668348954e-06, "loss": 0.0566, "step": 350 }, { "epoch": 0.05, "learning_rate": 6.398702235039654e-06, "loss": 0.0549, "step": 355 }, { "epoch": 0.05, "learning_rate": 6.488824801730353e-06, "loss": 0.0531, "step": 360 }, { "epoch": 0.05, "learning_rate": 6.578947368421053e-06, "loss": 0.0509, "step": 365 }, { "epoch": 0.05, "learning_rate": 6.669069935111752e-06, "loss": 0.0491, "step": 370 }, { "epoch": 0.05, "learning_rate": 6.759192501802451e-06, "loss": 0.0474, "step": 375 }, { "epoch": 0.05, "learning_rate": 6.849315068493151e-06, "loss": 0.0459, "step": 380 }, { "epoch": 0.06, "learning_rate": 6.93943763518385e-06, "loss": 0.0442, "step": 385 }, { "epoch": 0.06, "learning_rate": 7.029560201874549e-06, "loss": 0.043, "step": 390 }, { "epoch": 0.06, "learning_rate": 7.119682768565249e-06, "loss": 0.0408, "step": 395 }, { "epoch": 0.06, "learning_rate": 7.209805335255948e-06, "loss": 0.0399, "step": 400 }, { "epoch": 0.06, "learning_rate": 7.299927901946647e-06, "loss": 0.0386, "step": 405 }, { "epoch": 0.06, "learning_rate": 7.390050468637347e-06, "loss": 0.0369, "step": 410 }, { "epoch": 0.06, "learning_rate": 7.480173035328046e-06, "loss": 0.0356, "step": 415 }, { "epoch": 0.06, "learning_rate": 7.570295602018746e-06, "loss": 0.0349, "step": 420 }, { "epoch": 0.06, "learning_rate": 7.660418168709445e-06, "loss": 0.0334, "step": 425 }, { "epoch": 0.06, "learning_rate": 7.750540735400145e-06, "loss": 0.0325, "step": 430 }, { "epoch": 0.06, "learning_rate": 7.840663302090843e-06, "loss": 0.0317, "step": 435 }, { "epoch": 0.06, "learning_rate": 7.930785868781542e-06, "loss": 0.0305, "step": 440 }, { "epoch": 0.06, "learning_rate": 8.020908435472242e-06, "loss": 0.0296, "step": 445 }, { "epoch": 0.06, "learning_rate": 8.111031002162942e-06, "loss": 0.0287, "step": 450 }, { "epoch": 0.07, "learning_rate": 8.201153568853641e-06, "loss": 0.0277, "step": 455 }, { "epoch": 0.07, "learning_rate": 8.29127613554434e-06, "loss": 0.0272, "step": 460 }, { "epoch": 0.07, "learning_rate": 8.38139870223504e-06, "loss": 0.0264, "step": 465 }, { "epoch": 0.07, "learning_rate": 8.471521268925738e-06, "loss": 0.0254, "step": 470 }, { "epoch": 0.07, "learning_rate": 8.561643835616438e-06, "loss": 0.0248, "step": 475 }, { "epoch": 0.07, "learning_rate": 8.651766402307137e-06, "loss": 0.0238, "step": 480 }, { "epoch": 0.07, "learning_rate": 8.741888968997837e-06, "loss": 0.0235, "step": 485 }, { "epoch": 0.07, "learning_rate": 8.832011535688536e-06, "loss": 0.0226, "step": 490 }, { "epoch": 0.07, "learning_rate": 8.922134102379236e-06, "loss": 0.0218, "step": 495 }, { "epoch": 0.07, "learning_rate": 9.012256669069936e-06, "loss": 0.0216, "step": 500 }, { "epoch": 0.07, "learning_rate": 9.102379235760635e-06, "loss": 0.021, "step": 505 }, { "epoch": 0.07, "learning_rate": 9.192501802451333e-06, "loss": 0.0206, "step": 510 }, { "epoch": 0.07, "learning_rate": 9.282624369142033e-06, "loss": 0.0195, "step": 515 }, { "epoch": 0.08, "learning_rate": 9.372746935832732e-06, "loss": 0.0199, "step": 520 }, { "epoch": 0.08, "learning_rate": 9.462869502523432e-06, "loss": 0.0188, "step": 525 }, { "epoch": 0.08, "learning_rate": 9.552992069214131e-06, "loss": 0.0186, "step": 530 }, { "epoch": 0.08, "learning_rate": 9.643114635904831e-06, "loss": 0.018, "step": 535 }, { "epoch": 0.08, "learning_rate": 9.73323720259553e-06, "loss": 0.0179, "step": 540 }, { "epoch": 0.08, "learning_rate": 9.823359769286228e-06, "loss": 0.0172, "step": 545 }, { "epoch": 0.08, "learning_rate": 9.913482335976928e-06, "loss": 0.0167, "step": 550 }, { "epoch": 0.08, "learning_rate": 1.0003604902667628e-05, "loss": 0.0161, "step": 555 }, { "epoch": 0.08, "learning_rate": 1.0093727469358327e-05, "loss": 0.0163, "step": 560 }, { "epoch": 0.08, "learning_rate": 1.0183850036049027e-05, "loss": 0.0154, "step": 565 }, { "epoch": 0.08, "learning_rate": 1.0273972602739726e-05, "loss": 0.0151, "step": 570 }, { "epoch": 0.08, "learning_rate": 1.0364095169430426e-05, "loss": 0.0149, "step": 575 }, { "epoch": 0.08, "learning_rate": 1.0454217736121125e-05, "loss": 0.0148, "step": 580 }, { "epoch": 0.08, "learning_rate": 1.0544340302811823e-05, "loss": 0.0143, "step": 585 }, { "epoch": 0.09, "learning_rate": 1.0634462869502523e-05, "loss": 0.014, "step": 590 }, { "epoch": 0.09, "learning_rate": 1.0724585436193222e-05, "loss": 0.0135, "step": 595 }, { "epoch": 0.09, "learning_rate": 1.0814708002883922e-05, "loss": 0.0132, "step": 600 }, { "epoch": 0.09, "learning_rate": 1.0904830569574622e-05, "loss": 0.0134, "step": 605 }, { "epoch": 0.09, "learning_rate": 1.0994953136265321e-05, "loss": 0.013, "step": 610 }, { "epoch": 0.09, "learning_rate": 1.108507570295602e-05, "loss": 0.0124, "step": 615 }, { "epoch": 0.09, "learning_rate": 1.1175198269646719e-05, "loss": 0.0122, "step": 620 }, { "epoch": 0.09, "learning_rate": 1.1265320836337418e-05, "loss": 0.0124, "step": 625 }, { "epoch": 0.09, "learning_rate": 1.1355443403028118e-05, "loss": 0.012, "step": 630 }, { "epoch": 0.09, "learning_rate": 1.1445565969718817e-05, "loss": 0.0115, "step": 635 }, { "epoch": 0.09, "learning_rate": 1.1535688536409517e-05, "loss": 0.0116, "step": 640 }, { "epoch": 0.09, "learning_rate": 1.1625811103100217e-05, "loss": 0.0113, "step": 645 }, { "epoch": 0.09, "learning_rate": 1.1715933669790916e-05, "loss": 0.0107, "step": 650 }, { "epoch": 0.09, "learning_rate": 1.1806056236481616e-05, "loss": 0.0111, "step": 655 }, { "epoch": 0.1, "learning_rate": 1.1896178803172314e-05, "loss": 0.0105, "step": 660 }, { "epoch": 0.1, "learning_rate": 1.1986301369863013e-05, "loss": 0.0108, "step": 665 }, { "epoch": 0.1, "learning_rate": 1.2076423936553713e-05, "loss": 0.0103, "step": 670 }, { "epoch": 0.1, "learning_rate": 1.2166546503244412e-05, "loss": 0.0102, "step": 675 }, { "epoch": 0.1, "learning_rate": 1.2256669069935112e-05, "loss": 0.0098, "step": 680 }, { "epoch": 0.1, "learning_rate": 1.2346791636625811e-05, "loss": 0.0094, "step": 685 }, { "epoch": 0.1, "learning_rate": 1.2436914203316511e-05, "loss": 0.0094, "step": 690 }, { "epoch": 0.1, "learning_rate": 1.252703677000721e-05, "loss": 0.0096, "step": 695 }, { "epoch": 0.1, "learning_rate": 1.2617159336697909e-05, "loss": 0.009, "step": 700 }, { "epoch": 0.1, "learning_rate": 1.270728190338861e-05, "loss": 0.0091, "step": 705 }, { "epoch": 0.1, "learning_rate": 1.2797404470079308e-05, "loss": 0.0089, "step": 710 }, { "epoch": 0.1, "learning_rate": 1.2887527036770009e-05, "loss": 0.0086, "step": 715 }, { "epoch": 0.1, "learning_rate": 1.2977649603460707e-05, "loss": 0.008, "step": 720 }, { "epoch": 0.1, "learning_rate": 1.3067772170151406e-05, "loss": 0.0085, "step": 725 }, { "epoch": 0.11, "learning_rate": 1.3157894736842106e-05, "loss": 0.0078, "step": 730 }, { "epoch": 0.11, "learning_rate": 1.3248017303532806e-05, "loss": 0.008, "step": 735 }, { "epoch": 0.11, "learning_rate": 1.3338139870223503e-05, "loss": 0.0077, "step": 740 }, { "epoch": 0.11, "learning_rate": 1.3428262436914205e-05, "loss": 0.0081, "step": 745 }, { "epoch": 0.11, "learning_rate": 1.3518385003604903e-05, "loss": 0.0082, "step": 750 }, { "epoch": 0.11, "learning_rate": 1.3608507570295604e-05, "loss": 0.0082, "step": 755 }, { "epoch": 0.11, "learning_rate": 1.3698630136986302e-05, "loss": 0.008, "step": 760 }, { "epoch": 0.11, "learning_rate": 1.3788752703677001e-05, "loss": 0.0074, "step": 765 }, { "epoch": 0.11, "learning_rate": 1.38788752703677e-05, "loss": 0.0073, "step": 770 }, { "epoch": 0.11, "learning_rate": 1.39689978370584e-05, "loss": 0.0071, "step": 775 }, { "epoch": 0.11, "learning_rate": 1.4059120403749098e-05, "loss": 0.0073, "step": 780 }, { "epoch": 0.11, "learning_rate": 1.41492429704398e-05, "loss": 0.007, "step": 785 }, { "epoch": 0.11, "learning_rate": 1.4239365537130498e-05, "loss": 0.007, "step": 790 }, { "epoch": 0.11, "learning_rate": 1.4329488103821199e-05, "loss": 0.0065, "step": 795 }, { "epoch": 0.12, "learning_rate": 1.4419610670511897e-05, "loss": 0.0067, "step": 800 }, { "epoch": 0.12, "learning_rate": 1.4509733237202596e-05, "loss": 0.0074, "step": 805 }, { "epoch": 0.12, "learning_rate": 1.4599855803893294e-05, "loss": 0.0067, "step": 810 }, { "epoch": 0.12, "learning_rate": 1.4689978370583995e-05, "loss": 0.0063, "step": 815 }, { "epoch": 0.12, "learning_rate": 1.4780100937274693e-05, "loss": 0.0061, "step": 820 }, { "epoch": 0.12, "learning_rate": 1.4870223503965395e-05, "loss": 0.0065, "step": 825 }, { "epoch": 0.12, "learning_rate": 1.4960346070656092e-05, "loss": 0.0067, "step": 830 }, { "epoch": 0.12, "learning_rate": 1.5050468637346794e-05, "loss": 0.0069, "step": 835 }, { "epoch": 0.12, "learning_rate": 1.5140591204037492e-05, "loss": 0.0062, "step": 840 }, { "epoch": 0.12, "learning_rate": 1.5230713770728191e-05, "loss": 0.006, "step": 845 }, { "epoch": 0.12, "learning_rate": 1.532083633741889e-05, "loss": 0.0062, "step": 850 }, { "epoch": 0.12, "learning_rate": 1.541095890410959e-05, "loss": 0.0058, "step": 855 }, { "epoch": 0.12, "learning_rate": 1.550108147080029e-05, "loss": 0.0065, "step": 860 }, { "epoch": 0.12, "learning_rate": 1.5591204037490988e-05, "loss": 0.006, "step": 865 }, { "epoch": 0.13, "learning_rate": 1.5681326604181686e-05, "loss": 0.0055, "step": 870 }, { "epoch": 0.13, "learning_rate": 1.5771449170872387e-05, "loss": 0.0057, "step": 875 }, { "epoch": 0.13, "learning_rate": 1.5861571737563085e-05, "loss": 0.0057, "step": 880 }, { "epoch": 0.13, "learning_rate": 1.5951694304253786e-05, "loss": 0.0058, "step": 885 }, { "epoch": 0.13, "learning_rate": 1.6041816870944484e-05, "loss": 0.005, "step": 890 }, { "epoch": 0.13, "learning_rate": 1.6131939437635185e-05, "loss": 0.0054, "step": 895 }, { "epoch": 0.13, "learning_rate": 1.6222062004325883e-05, "loss": 0.0058, "step": 900 }, { "epoch": 0.13, "learning_rate": 1.6312184571016584e-05, "loss": 0.0049, "step": 905 }, { "epoch": 0.13, "learning_rate": 1.6402307137707282e-05, "loss": 0.005, "step": 910 }, { "epoch": 0.13, "learning_rate": 1.6492429704397984e-05, "loss": 0.005, "step": 915 }, { "epoch": 0.13, "learning_rate": 1.658255227108868e-05, "loss": 0.0054, "step": 920 }, { "epoch": 0.13, "learning_rate": 1.6672674837779383e-05, "loss": 0.0057, "step": 925 }, { "epoch": 0.13, "learning_rate": 1.676279740447008e-05, "loss": 0.0051, "step": 930 }, { "epoch": 0.13, "learning_rate": 1.685291997116078e-05, "loss": 0.0051, "step": 935 }, { "epoch": 0.14, "learning_rate": 1.6943042537851476e-05, "loss": 0.0049, "step": 940 }, { "epoch": 0.14, "learning_rate": 1.7033165104542178e-05, "loss": 0.0048, "step": 945 }, { "epoch": 0.14, "learning_rate": 1.7123287671232875e-05, "loss": 0.0049, "step": 950 }, { "epoch": 0.14, "learning_rate": 1.7213410237923577e-05, "loss": 0.0049, "step": 955 }, { "epoch": 0.14, "learning_rate": 1.7303532804614275e-05, "loss": 0.0047, "step": 960 }, { "epoch": 0.14, "learning_rate": 1.7393655371304976e-05, "loss": 0.0047, "step": 965 }, { "epoch": 0.14, "learning_rate": 1.7483777937995674e-05, "loss": 0.0044, "step": 970 }, { "epoch": 0.14, "learning_rate": 1.7573900504686375e-05, "loss": 0.0043, "step": 975 }, { "epoch": 0.14, "learning_rate": 1.7664023071377073e-05, "loss": 0.0045, "step": 980 }, { "epoch": 0.14, "learning_rate": 1.7754145638067774e-05, "loss": 0.0043, "step": 985 }, { "epoch": 0.14, "learning_rate": 1.7844268204758472e-05, "loss": 0.0045, "step": 990 }, { "epoch": 0.14, "learning_rate": 1.7934390771449173e-05, "loss": 0.0045, "step": 995 }, { "epoch": 0.14, "learning_rate": 1.802451333813987e-05, "loss": 0.0042, "step": 1000 }, { "epoch": 0.14, "learning_rate": 1.811463590483057e-05, "loss": 0.0045, "step": 1005 }, { "epoch": 0.15, "learning_rate": 1.820475847152127e-05, "loss": 0.0044, "step": 1010 }, { "epoch": 0.15, "learning_rate": 1.8294881038211968e-05, "loss": 0.0046, "step": 1015 }, { "epoch": 0.15, "learning_rate": 1.8385003604902666e-05, "loss": 0.0048, "step": 1020 }, { "epoch": 0.15, "learning_rate": 1.8475126171593367e-05, "loss": 0.0042, "step": 1025 }, { "epoch": 0.15, "learning_rate": 1.8565248738284065e-05, "loss": 0.0039, "step": 1030 }, { "epoch": 0.15, "learning_rate": 1.8655371304974767e-05, "loss": 0.0047, "step": 1035 }, { "epoch": 0.15, "learning_rate": 1.8745493871665464e-05, "loss": 0.0041, "step": 1040 }, { "epoch": 0.15, "learning_rate": 1.8835616438356166e-05, "loss": 0.0039, "step": 1045 }, { "epoch": 0.15, "learning_rate": 1.8925739005046864e-05, "loss": 0.0038, "step": 1050 }, { "epoch": 0.15, "learning_rate": 1.9015861571737565e-05, "loss": 0.0045, "step": 1055 }, { "epoch": 0.15, "learning_rate": 1.9105984138428263e-05, "loss": 0.0049, "step": 1060 }, { "epoch": 0.15, "learning_rate": 1.9196106705118964e-05, "loss": 0.0046, "step": 1065 }, { "epoch": 0.15, "learning_rate": 1.9286229271809662e-05, "loss": 0.0036, "step": 1070 }, { "epoch": 0.16, "learning_rate": 1.9376351838500363e-05, "loss": 0.0039, "step": 1075 }, { "epoch": 0.16, "learning_rate": 1.946647440519106e-05, "loss": 0.0041, "step": 1080 }, { "epoch": 0.16, "learning_rate": 1.955659697188176e-05, "loss": 0.0047, "step": 1085 }, { "epoch": 0.16, "learning_rate": 1.9646719538572457e-05, "loss": 0.004, "step": 1090 }, { "epoch": 0.16, "learning_rate": 1.9736842105263158e-05, "loss": 0.0035, "step": 1095 }, { "epoch": 0.16, "learning_rate": 1.9826964671953856e-05, "loss": 0.0042, "step": 1100 }, { "epoch": 0.16, "learning_rate": 1.9917087238644557e-05, "loss": 0.0037, "step": 1105 }, { "epoch": 0.16, "learning_rate": 2.0007209805335255e-05, "loss": 0.0036, "step": 1110 }, { "epoch": 0.16, "learning_rate": 2.0097332372025956e-05, "loss": 0.0038, "step": 1115 }, { "epoch": 0.16, "learning_rate": 2.0187454938716654e-05, "loss": 0.0036, "step": 1120 }, { "epoch": 0.16, "learning_rate": 2.0277577505407356e-05, "loss": 0.0037, "step": 1125 }, { "epoch": 0.16, "learning_rate": 2.0367700072098053e-05, "loss": 0.0038, "step": 1130 }, { "epoch": 0.16, "learning_rate": 2.0457822638788755e-05, "loss": 0.0039, "step": 1135 }, { "epoch": 0.16, "learning_rate": 2.0547945205479453e-05, "loss": 0.0038, "step": 1140 }, { "epoch": 0.17, "learning_rate": 2.0638067772170154e-05, "loss": 0.0034, "step": 1145 }, { "epoch": 0.17, "learning_rate": 2.0728190338860852e-05, "loss": 0.0039, "step": 1150 }, { "epoch": 0.17, "learning_rate": 2.081831290555155e-05, "loss": 0.0036, "step": 1155 }, { "epoch": 0.17, "learning_rate": 2.090843547224225e-05, "loss": 0.0031, "step": 1160 }, { "epoch": 0.17, "learning_rate": 2.099855803893295e-05, "loss": 0.0038, "step": 1165 }, { "epoch": 0.17, "learning_rate": 2.1088680605623647e-05, "loss": 0.0039, "step": 1170 }, { "epoch": 0.17, "learning_rate": 2.1178803172314348e-05, "loss": 0.0037, "step": 1175 }, { "epoch": 0.17, "learning_rate": 2.1268925739005046e-05, "loss": 0.0035, "step": 1180 }, { "epoch": 0.17, "learning_rate": 2.1359048305695747e-05, "loss": 0.0032, "step": 1185 }, { "epoch": 0.17, "learning_rate": 2.1449170872386445e-05, "loss": 0.0037, "step": 1190 }, { "epoch": 0.17, "learning_rate": 2.1539293439077146e-05, "loss": 0.0037, "step": 1195 }, { "epoch": 0.17, "learning_rate": 2.1629416005767844e-05, "loss": 0.0034, "step": 1200 }, { "epoch": 0.17, "learning_rate": 2.1719538572458545e-05, "loss": 0.0047, "step": 1205 }, { "epoch": 0.17, "learning_rate": 2.1809661139149243e-05, "loss": 0.0036, "step": 1210 }, { "epoch": 0.18, "learning_rate": 2.1899783705839945e-05, "loss": 0.0039, "step": 1215 }, { "epoch": 0.18, "learning_rate": 2.1989906272530642e-05, "loss": 0.0031, "step": 1220 }, { "epoch": 0.18, "learning_rate": 2.2080028839221344e-05, "loss": 0.0039, "step": 1225 }, { "epoch": 0.18, "learning_rate": 2.217015140591204e-05, "loss": 0.0033, "step": 1230 }, { "epoch": 0.18, "learning_rate": 2.226027397260274e-05, "loss": 0.0039, "step": 1235 }, { "epoch": 0.18, "learning_rate": 2.2350396539293437e-05, "loss": 0.0038, "step": 1240 }, { "epoch": 0.18, "learning_rate": 2.244051910598414e-05, "loss": 0.0037, "step": 1245 }, { "epoch": 0.18, "learning_rate": 2.2530641672674837e-05, "loss": 0.0034, "step": 1250 }, { "epoch": 0.18, "learning_rate": 2.2620764239365538e-05, "loss": 0.0035, "step": 1255 }, { "epoch": 0.18, "learning_rate": 2.2710886806056236e-05, "loss": 0.0037, "step": 1260 }, { "epoch": 0.18, "learning_rate": 2.2801009372746937e-05, "loss": 0.0035, "step": 1265 }, { "epoch": 0.18, "learning_rate": 2.2891131939437635e-05, "loss": 0.003, "step": 1270 }, { "epoch": 0.18, "learning_rate": 2.2981254506128336e-05, "loss": 0.0033, "step": 1275 }, { "epoch": 0.18, "learning_rate": 2.3071377072819034e-05, "loss": 0.0031, "step": 1280 }, { "epoch": 0.19, "learning_rate": 2.3161499639509735e-05, "loss": 0.0026, "step": 1285 }, { "epoch": 0.19, "learning_rate": 2.3251622206200433e-05, "loss": 0.0033, "step": 1290 }, { "epoch": 0.19, "learning_rate": 2.3341744772891134e-05, "loss": 0.0035, "step": 1295 }, { "epoch": 0.19, "learning_rate": 2.3431867339581832e-05, "loss": 0.0032, "step": 1300 }, { "epoch": 0.19, "learning_rate": 2.352198990627253e-05, "loss": 0.0033, "step": 1305 }, { "epoch": 0.19, "learning_rate": 2.361211247296323e-05, "loss": 0.0034, "step": 1310 }, { "epoch": 0.19, "learning_rate": 2.370223503965393e-05, "loss": 0.0035, "step": 1315 }, { "epoch": 0.19, "learning_rate": 2.3792357606344627e-05, "loss": 0.0031, "step": 1320 }, { "epoch": 0.19, "learning_rate": 2.388248017303533e-05, "loss": 0.0032, "step": 1325 }, { "epoch": 0.19, "learning_rate": 2.3972602739726026e-05, "loss": 0.0037, "step": 1330 }, { "epoch": 0.19, "learning_rate": 2.4062725306416728e-05, "loss": 0.0033, "step": 1335 }, { "epoch": 0.19, "learning_rate": 2.4152847873107426e-05, "loss": 0.0025, "step": 1340 }, { "epoch": 0.19, "learning_rate": 2.4242970439798127e-05, "loss": 0.0028, "step": 1345 }, { "epoch": 0.19, "learning_rate": 2.4333093006488825e-05, "loss": 0.0031, "step": 1350 }, { "epoch": 0.2, "learning_rate": 2.4423215573179526e-05, "loss": 0.0031, "step": 1355 }, { "epoch": 0.2, "learning_rate": 2.4513338139870224e-05, "loss": 0.0024, "step": 1360 }, { "epoch": 0.2, "learning_rate": 2.4603460706560925e-05, "loss": 0.0032, "step": 1365 }, { "epoch": 0.2, "learning_rate": 2.4693583273251623e-05, "loss": 0.0033, "step": 1370 }, { "epoch": 0.2, "learning_rate": 2.4783705839942324e-05, "loss": 0.0028, "step": 1375 }, { "epoch": 0.2, "learning_rate": 2.4873828406633022e-05, "loss": 0.0033, "step": 1380 }, { "epoch": 0.2, "learning_rate": 2.496395097332372e-05, "loss": 0.0035, "step": 1385 }, { "epoch": 0.2, "learning_rate": 2.505407354001442e-05, "loss": 0.0031, "step": 1390 }, { "epoch": 0.2, "learning_rate": 2.514419610670512e-05, "loss": 0.0036, "step": 1395 }, { "epoch": 0.2, "learning_rate": 2.5234318673395817e-05, "loss": 0.0028, "step": 1400 }, { "epoch": 0.2, "learning_rate": 2.5324441240086522e-05, "loss": 0.0027, "step": 1405 }, { "epoch": 0.2, "learning_rate": 2.541456380677722e-05, "loss": 0.0033, "step": 1410 }, { "epoch": 0.2, "learning_rate": 2.5504686373467917e-05, "loss": 0.0034, "step": 1415 }, { "epoch": 0.2, "learning_rate": 2.5594808940158615e-05, "loss": 0.0026, "step": 1420 }, { "epoch": 0.21, "learning_rate": 2.568493150684932e-05, "loss": 0.0034, "step": 1425 }, { "epoch": 0.21, "learning_rate": 2.5775054073540018e-05, "loss": 0.0027, "step": 1430 }, { "epoch": 0.21, "learning_rate": 2.5865176640230716e-05, "loss": 0.0035, "step": 1435 }, { "epoch": 0.21, "learning_rate": 2.5955299206921414e-05, "loss": 0.0025, "step": 1440 }, { "epoch": 0.21, "learning_rate": 2.6045421773612115e-05, "loss": 0.0037, "step": 1445 }, { "epoch": 0.21, "learning_rate": 2.6135544340302813e-05, "loss": 0.0031, "step": 1450 }, { "epoch": 0.21, "learning_rate": 2.622566690699351e-05, "loss": 0.0034, "step": 1455 }, { "epoch": 0.21, "learning_rate": 2.6315789473684212e-05, "loss": 0.003, "step": 1460 }, { "epoch": 0.21, "learning_rate": 2.6405912040374913e-05, "loss": 0.0033, "step": 1465 }, { "epoch": 0.21, "learning_rate": 2.649603460706561e-05, "loss": 0.0028, "step": 1470 }, { "epoch": 0.21, "learning_rate": 2.658615717375631e-05, "loss": 0.0031, "step": 1475 }, { "epoch": 0.21, "learning_rate": 2.6676279740447007e-05, "loss": 0.0027, "step": 1480 }, { "epoch": 0.21, "learning_rate": 2.676640230713771e-05, "loss": 0.0029, "step": 1485 }, { "epoch": 0.21, "learning_rate": 2.685652487382841e-05, "loss": 0.0032, "step": 1490 }, { "epoch": 0.22, "learning_rate": 2.6946647440519107e-05, "loss": 0.0029, "step": 1495 }, { "epoch": 0.22, "learning_rate": 2.7036770007209805e-05, "loss": 0.0036, "step": 1500 }, { "epoch": 0.22, "learning_rate": 2.712689257390051e-05, "loss": 0.0021, "step": 1505 }, { "epoch": 0.22, "learning_rate": 2.7217015140591208e-05, "loss": 0.003, "step": 1510 }, { "epoch": 0.22, "learning_rate": 2.7307137707281906e-05, "loss": 0.0033, "step": 1515 }, { "epoch": 0.22, "learning_rate": 2.7397260273972603e-05, "loss": 0.0033, "step": 1520 }, { "epoch": 0.22, "learning_rate": 2.7487382840663305e-05, "loss": 0.0028, "step": 1525 }, { "epoch": 0.22, "learning_rate": 2.7577505407354003e-05, "loss": 0.0031, "step": 1530 }, { "epoch": 0.22, "learning_rate": 2.76676279740447e-05, "loss": 0.0032, "step": 1535 }, { "epoch": 0.22, "learning_rate": 2.77577505407354e-05, "loss": 0.0026, "step": 1540 }, { "epoch": 0.22, "learning_rate": 2.7847873107426103e-05, "loss": 0.0036, "step": 1545 }, { "epoch": 0.22, "learning_rate": 2.79379956741168e-05, "loss": 0.0029, "step": 1550 }, { "epoch": 0.22, "learning_rate": 2.80281182408075e-05, "loss": 0.0031, "step": 1555 }, { "epoch": 0.23, "learning_rate": 2.8118240807498197e-05, "loss": 0.0023, "step": 1560 }, { "epoch": 0.23, "learning_rate": 2.82083633741889e-05, "loss": 0.0032, "step": 1565 }, { "epoch": 0.23, "learning_rate": 2.82984859408796e-05, "loss": 0.0029, "step": 1570 }, { "epoch": 0.23, "learning_rate": 2.8388608507570297e-05, "loss": 0.0035, "step": 1575 }, { "epoch": 0.23, "learning_rate": 2.8478731074260995e-05, "loss": 0.0025, "step": 1580 }, { "epoch": 0.23, "learning_rate": 2.85688536409517e-05, "loss": 0.0028, "step": 1585 }, { "epoch": 0.23, "learning_rate": 2.8658976207642398e-05, "loss": 0.0033, "step": 1590 }, { "epoch": 0.23, "learning_rate": 2.8749098774333095e-05, "loss": 0.003, "step": 1595 }, { "epoch": 0.23, "learning_rate": 2.8839221341023793e-05, "loss": 0.0031, "step": 1600 }, { "epoch": 0.23, "learning_rate": 2.8929343907714495e-05, "loss": 0.0033, "step": 1605 }, { "epoch": 0.23, "learning_rate": 2.9019466474405192e-05, "loss": 0.0029, "step": 1610 }, { "epoch": 0.23, "learning_rate": 2.910958904109589e-05, "loss": 0.0033, "step": 1615 }, { "epoch": 0.23, "learning_rate": 2.9199711607786588e-05, "loss": 0.0033, "step": 1620 }, { "epoch": 0.23, "learning_rate": 2.9289834174477293e-05, "loss": 0.0025, "step": 1625 }, { "epoch": 0.24, "learning_rate": 2.937995674116799e-05, "loss": 0.0023, "step": 1630 }, { "epoch": 0.24, "learning_rate": 2.947007930785869e-05, "loss": 0.0035, "step": 1635 }, { "epoch": 0.24, "learning_rate": 2.9560201874549387e-05, "loss": 0.0029, "step": 1640 }, { "epoch": 0.24, "learning_rate": 2.965032444124009e-05, "loss": 0.0028, "step": 1645 }, { "epoch": 0.24, "learning_rate": 2.974044700793079e-05, "loss": 0.0031, "step": 1650 }, { "epoch": 0.24, "learning_rate": 2.9830569574621487e-05, "loss": 0.003, "step": 1655 }, { "epoch": 0.24, "learning_rate": 2.9920692141312185e-05, "loss": 0.0028, "step": 1660 }, { "epoch": 0.24, "learning_rate": 3.001081470800289e-05, "loss": 0.0029, "step": 1665 }, { "epoch": 0.24, "learning_rate": 3.0100937274693587e-05, "loss": 0.0027, "step": 1670 }, { "epoch": 0.24, "learning_rate": 3.0191059841384285e-05, "loss": 0.0026, "step": 1675 }, { "epoch": 0.24, "learning_rate": 3.0281182408074983e-05, "loss": 0.0024, "step": 1680 }, { "epoch": 0.24, "learning_rate": 3.0371304974765684e-05, "loss": 0.0024, "step": 1685 }, { "epoch": 0.24, "learning_rate": 3.0461427541456382e-05, "loss": 0.0023, "step": 1690 }, { "epoch": 0.24, "learning_rate": 3.0551550108147084e-05, "loss": 0.0026, "step": 1695 }, { "epoch": 0.25, "learning_rate": 3.064167267483778e-05, "loss": 0.0029, "step": 1700 }, { "epoch": 0.25, "learning_rate": 3.073179524152848e-05, "loss": 0.0031, "step": 1705 }, { "epoch": 0.25, "learning_rate": 3.082191780821918e-05, "loss": 0.0025, "step": 1710 }, { "epoch": 0.25, "learning_rate": 3.091204037490988e-05, "loss": 0.0034, "step": 1715 }, { "epoch": 0.25, "learning_rate": 3.100216294160058e-05, "loss": 0.0033, "step": 1720 }, { "epoch": 0.25, "learning_rate": 3.109228550829128e-05, "loss": 0.0028, "step": 1725 }, { "epoch": 0.25, "learning_rate": 3.1182408074981976e-05, "loss": 0.0025, "step": 1730 }, { "epoch": 0.25, "learning_rate": 3.1272530641672673e-05, "loss": 0.0023, "step": 1735 }, { "epoch": 0.25, "learning_rate": 3.136265320836337e-05, "loss": 0.0027, "step": 1740 }, { "epoch": 0.25, "learning_rate": 3.1452775775054076e-05, "loss": 0.0023, "step": 1745 }, { "epoch": 0.25, "learning_rate": 3.1542898341744774e-05, "loss": 0.0022, "step": 1750 }, { "epoch": 0.25, "learning_rate": 3.163302090843547e-05, "loss": 0.0025, "step": 1755 }, { "epoch": 0.25, "learning_rate": 3.172314347512617e-05, "loss": 0.0027, "step": 1760 }, { "epoch": 0.25, "learning_rate": 3.1813266041816874e-05, "loss": 0.0024, "step": 1765 }, { "epoch": 0.26, "learning_rate": 3.190338860850757e-05, "loss": 0.0022, "step": 1770 }, { "epoch": 0.26, "learning_rate": 3.199351117519827e-05, "loss": 0.0027, "step": 1775 }, { "epoch": 0.26, "learning_rate": 3.208363374188897e-05, "loss": 0.0025, "step": 1780 }, { "epoch": 0.26, "learning_rate": 3.217375630857967e-05, "loss": 0.0032, "step": 1785 }, { "epoch": 0.26, "learning_rate": 3.226387887527037e-05, "loss": 0.0022, "step": 1790 }, { "epoch": 0.26, "learning_rate": 3.235400144196107e-05, "loss": 0.0026, "step": 1795 }, { "epoch": 0.26, "learning_rate": 3.2444124008651766e-05, "loss": 0.0028, "step": 1800 }, { "epoch": 0.26, "learning_rate": 3.253424657534247e-05, "loss": 0.0026, "step": 1805 }, { "epoch": 0.26, "learning_rate": 3.262436914203317e-05, "loss": 0.0028, "step": 1810 }, { "epoch": 0.26, "learning_rate": 3.271449170872387e-05, "loss": 0.0025, "step": 1815 }, { "epoch": 0.26, "learning_rate": 3.2804614275414565e-05, "loss": 0.0022, "step": 1820 }, { "epoch": 0.26, "learning_rate": 3.289473684210527e-05, "loss": 0.003, "step": 1825 }, { "epoch": 0.26, "learning_rate": 3.298485940879597e-05, "loss": 0.0033, "step": 1830 }, { "epoch": 0.26, "learning_rate": 3.3074981975486665e-05, "loss": 0.0029, "step": 1835 }, { "epoch": 0.27, "learning_rate": 3.316510454217736e-05, "loss": 0.0024, "step": 1840 }, { "epoch": 0.27, "learning_rate": 3.325522710886807e-05, "loss": 0.0021, "step": 1845 }, { "epoch": 0.27, "learning_rate": 3.3345349675558765e-05, "loss": 0.0022, "step": 1850 }, { "epoch": 0.27, "learning_rate": 3.343547224224946e-05, "loss": 0.003, "step": 1855 }, { "epoch": 0.27, "learning_rate": 3.352559480894016e-05, "loss": 0.0033, "step": 1860 }, { "epoch": 0.27, "learning_rate": 3.361571737563086e-05, "loss": 0.0019, "step": 1865 }, { "epoch": 0.27, "learning_rate": 3.370583994232156e-05, "loss": 0.0024, "step": 1870 }, { "epoch": 0.27, "learning_rate": 3.3795962509012255e-05, "loss": 0.0018, "step": 1875 }, { "epoch": 0.27, "learning_rate": 3.388608507570295e-05, "loss": 0.0029, "step": 1880 }, { "epoch": 0.27, "learning_rate": 3.397620764239366e-05, "loss": 0.0034, "step": 1885 }, { "epoch": 0.27, "learning_rate": 3.4066330209084355e-05, "loss": 0.0024, "step": 1890 }, { "epoch": 0.27, "learning_rate": 3.415645277577505e-05, "loss": 0.0034, "step": 1895 }, { "epoch": 0.27, "learning_rate": 3.424657534246575e-05, "loss": 0.0035, "step": 1900 }, { "epoch": 0.27, "learning_rate": 3.4336697909156456e-05, "loss": 0.0026, "step": 1905 }, { "epoch": 0.28, "learning_rate": 3.4426820475847154e-05, "loss": 0.0033, "step": 1910 }, { "epoch": 0.28, "learning_rate": 3.451694304253785e-05, "loss": 0.0025, "step": 1915 }, { "epoch": 0.28, "learning_rate": 3.460706560922855e-05, "loss": 0.0024, "step": 1920 }, { "epoch": 0.28, "learning_rate": 3.4697188175919254e-05, "loss": 0.0025, "step": 1925 }, { "epoch": 0.28, "learning_rate": 3.478731074260995e-05, "loss": 0.0026, "step": 1930 }, { "epoch": 0.28, "learning_rate": 3.487743330930065e-05, "loss": 0.0023, "step": 1935 }, { "epoch": 0.28, "learning_rate": 3.496755587599135e-05, "loss": 0.0027, "step": 1940 }, { "epoch": 0.28, "learning_rate": 3.505767844268205e-05, "loss": 0.0025, "step": 1945 }, { "epoch": 0.28, "learning_rate": 3.514780100937275e-05, "loss": 0.0019, "step": 1950 }, { "epoch": 0.28, "learning_rate": 3.523792357606345e-05, "loss": 0.0028, "step": 1955 }, { "epoch": 0.28, "learning_rate": 3.5328046142754146e-05, "loss": 0.0017, "step": 1960 }, { "epoch": 0.28, "learning_rate": 3.541816870944485e-05, "loss": 0.0025, "step": 1965 }, { "epoch": 0.28, "learning_rate": 3.550829127613555e-05, "loss": 0.0022, "step": 1970 }, { "epoch": 0.28, "learning_rate": 3.5598413842826246e-05, "loss": 0.0021, "step": 1975 }, { "epoch": 0.29, "learning_rate": 3.5688536409516944e-05, "loss": 0.0036, "step": 1980 }, { "epoch": 0.29, "learning_rate": 3.577865897620765e-05, "loss": 0.003, "step": 1985 }, { "epoch": 0.29, "learning_rate": 3.586878154289835e-05, "loss": 0.0021, "step": 1990 }, { "epoch": 0.29, "learning_rate": 3.5958904109589045e-05, "loss": 0.0029, "step": 1995 }, { "epoch": 0.29, "learning_rate": 3.604902667627974e-05, "loss": 0.0033, "step": 2000 }, { "epoch": 0.29, "learning_rate": 3.613914924297044e-05, "loss": 0.0028, "step": 2005 }, { "epoch": 0.29, "learning_rate": 3.622927180966114e-05, "loss": 0.0021, "step": 2010 }, { "epoch": 0.29, "learning_rate": 3.631939437635184e-05, "loss": 0.002, "step": 2015 }, { "epoch": 0.29, "learning_rate": 3.640951694304254e-05, "loss": 0.0024, "step": 2020 }, { "epoch": 0.29, "learning_rate": 3.649963950973324e-05, "loss": 0.0025, "step": 2025 }, { "epoch": 0.29, "learning_rate": 3.6589762076423937e-05, "loss": 0.0031, "step": 2030 }, { "epoch": 0.29, "learning_rate": 3.6679884643114634e-05, "loss": 0.0033, "step": 2035 }, { "epoch": 0.29, "learning_rate": 3.677000720980533e-05, "loss": 0.0027, "step": 2040 }, { "epoch": 0.29, "learning_rate": 3.686012977649604e-05, "loss": 0.0027, "step": 2045 }, { "epoch": 0.3, "learning_rate": 3.6950252343186735e-05, "loss": 0.0026, "step": 2050 }, { "epoch": 0.3, "learning_rate": 3.704037490987743e-05, "loss": 0.0024, "step": 2055 }, { "epoch": 0.3, "learning_rate": 3.713049747656813e-05, "loss": 0.0024, "step": 2060 }, { "epoch": 0.3, "learning_rate": 3.7220620043258835e-05, "loss": 0.0031, "step": 2065 }, { "epoch": 0.3, "learning_rate": 3.731074260994953e-05, "loss": 0.0028, "step": 2070 }, { "epoch": 0.3, "learning_rate": 3.740086517664023e-05, "loss": 0.0024, "step": 2075 }, { "epoch": 0.3, "learning_rate": 3.749098774333093e-05, "loss": 0.0027, "step": 2080 }, { "epoch": 0.3, "learning_rate": 3.7581110310021634e-05, "loss": 0.0028, "step": 2085 }, { "epoch": 0.3, "learning_rate": 3.767123287671233e-05, "loss": 0.0026, "step": 2090 }, { "epoch": 0.3, "learning_rate": 3.776135544340303e-05, "loss": 0.0022, "step": 2095 }, { "epoch": 0.3, "learning_rate": 3.785147801009373e-05, "loss": 0.0023, "step": 2100 }, { "epoch": 0.3, "learning_rate": 3.794160057678443e-05, "loss": 0.003, "step": 2105 }, { "epoch": 0.3, "learning_rate": 3.803172314347513e-05, "loss": 0.0022, "step": 2110 }, { "epoch": 0.31, "learning_rate": 3.812184571016583e-05, "loss": 0.0024, "step": 2115 }, { "epoch": 0.31, "learning_rate": 3.8211968276856526e-05, "loss": 0.0028, "step": 2120 }, { "epoch": 0.31, "learning_rate": 3.830209084354723e-05, "loss": 0.0027, "step": 2125 }, { "epoch": 0.31, "learning_rate": 3.839221341023793e-05, "loss": 0.0024, "step": 2130 }, { "epoch": 0.31, "learning_rate": 3.8482335976928626e-05, "loss": 0.0023, "step": 2135 }, { "epoch": 0.31, "learning_rate": 3.8572458543619324e-05, "loss": 0.0027, "step": 2140 }, { "epoch": 0.31, "learning_rate": 3.866258111031003e-05, "loss": 0.0029, "step": 2145 }, { "epoch": 0.31, "learning_rate": 3.8752703677000726e-05, "loss": 0.002, "step": 2150 }, { "epoch": 0.31, "learning_rate": 3.8842826243691424e-05, "loss": 0.0018, "step": 2155 }, { "epoch": 0.31, "learning_rate": 3.893294881038212e-05, "loss": 0.0017, "step": 2160 }, { "epoch": 0.31, "learning_rate": 3.902307137707282e-05, "loss": 0.003, "step": 2165 }, { "epoch": 0.31, "learning_rate": 3.911319394376352e-05, "loss": 0.0021, "step": 2170 }, { "epoch": 0.31, "learning_rate": 3.9203316510454216e-05, "loss": 0.002, "step": 2175 }, { "epoch": 0.31, "learning_rate": 3.9293439077144914e-05, "loss": 0.0023, "step": 2180 }, { "epoch": 0.32, "learning_rate": 3.938356164383562e-05, "loss": 0.003, "step": 2185 }, { "epoch": 0.32, "learning_rate": 3.9473684210526316e-05, "loss": 0.0022, "step": 2190 }, { "epoch": 0.32, "learning_rate": 3.9563806777217014e-05, "loss": 0.0023, "step": 2195 }, { "epoch": 0.32, "learning_rate": 3.965392934390771e-05, "loss": 0.0034, "step": 2200 }, { "epoch": 0.32, "learning_rate": 3.974405191059842e-05, "loss": 0.0019, "step": 2205 }, { "epoch": 0.32, "learning_rate": 3.9834174477289115e-05, "loss": 0.0019, "step": 2210 }, { "epoch": 0.32, "learning_rate": 3.992429704397981e-05, "loss": 0.0025, "step": 2215 }, { "epoch": 0.32, "learning_rate": 4.001441961067051e-05, "loss": 0.0026, "step": 2220 }, { "epoch": 0.32, "learning_rate": 4.0104542177361215e-05, "loss": 0.0024, "step": 2225 }, { "epoch": 0.32, "learning_rate": 4.019466474405191e-05, "loss": 0.0024, "step": 2230 }, { "epoch": 0.32, "learning_rate": 4.028478731074261e-05, "loss": 0.0021, "step": 2235 }, { "epoch": 0.32, "learning_rate": 4.037490987743331e-05, "loss": 0.0025, "step": 2240 }, { "epoch": 0.32, "learning_rate": 4.046503244412401e-05, "loss": 0.0021, "step": 2245 }, { "epoch": 0.32, "learning_rate": 4.055515501081471e-05, "loss": 0.0029, "step": 2250 }, { "epoch": 0.33, "learning_rate": 4.064527757750541e-05, "loss": 0.0017, "step": 2255 }, { "epoch": 0.33, "learning_rate": 4.073540014419611e-05, "loss": 0.0026, "step": 2260 }, { "epoch": 0.33, "learning_rate": 4.082552271088681e-05, "loss": 0.0032, "step": 2265 }, { "epoch": 0.33, "learning_rate": 4.091564527757751e-05, "loss": 0.0026, "step": 2270 }, { "epoch": 0.33, "learning_rate": 4.100576784426821e-05, "loss": 0.0029, "step": 2275 }, { "epoch": 0.33, "learning_rate": 4.1095890410958905e-05, "loss": 0.0027, "step": 2280 }, { "epoch": 0.33, "learning_rate": 4.118601297764961e-05, "loss": 0.0022, "step": 2285 }, { "epoch": 0.33, "learning_rate": 4.127613554434031e-05, "loss": 0.0019, "step": 2290 }, { "epoch": 0.33, "learning_rate": 4.1366258111031006e-05, "loss": 0.0024, "step": 2295 }, { "epoch": 0.33, "learning_rate": 4.1456380677721704e-05, "loss": 0.0016, "step": 2300 }, { "epoch": 0.33, "learning_rate": 4.15465032444124e-05, "loss": 0.0017, "step": 2305 }, { "epoch": 0.33, "learning_rate": 4.16366258111031e-05, "loss": 0.0023, "step": 2310 }, { "epoch": 0.33, "learning_rate": 4.1726748377793804e-05, "loss": 0.0013, "step": 2315 }, { "epoch": 0.33, "learning_rate": 4.18168709444845e-05, "loss": 0.0025, "step": 2320 }, { "epoch": 0.34, "learning_rate": 4.19069935111752e-05, "loss": 0.0027, "step": 2325 }, { "epoch": 0.34, "learning_rate": 4.19971160778659e-05, "loss": 0.0024, "step": 2330 }, { "epoch": 0.34, "learning_rate": 4.2087238644556596e-05, "loss": 0.0024, "step": 2335 }, { "epoch": 0.34, "learning_rate": 4.217736121124729e-05, "loss": 0.0016, "step": 2340 }, { "epoch": 0.34, "learning_rate": 4.2267483777938e-05, "loss": 0.0021, "step": 2345 }, { "epoch": 0.34, "learning_rate": 4.2357606344628696e-05, "loss": 0.0022, "step": 2350 }, { "epoch": 0.34, "learning_rate": 4.2447728911319394e-05, "loss": 0.0028, "step": 2355 }, { "epoch": 0.34, "learning_rate": 4.253785147801009e-05, "loss": 0.0026, "step": 2360 }, { "epoch": 0.34, "learning_rate": 4.2627974044700796e-05, "loss": 0.0019, "step": 2365 }, { "epoch": 0.34, "learning_rate": 4.2718096611391494e-05, "loss": 0.0019, "step": 2370 }, { "epoch": 0.34, "learning_rate": 4.280821917808219e-05, "loss": 0.003, "step": 2375 }, { "epoch": 0.34, "learning_rate": 4.289834174477289e-05, "loss": 0.0024, "step": 2380 }, { "epoch": 0.34, "learning_rate": 4.2988464311463595e-05, "loss": 0.0021, "step": 2385 }, { "epoch": 0.34, "learning_rate": 4.307858687815429e-05, "loss": 0.0016, "step": 2390 }, { "epoch": 0.35, "learning_rate": 4.316870944484499e-05, "loss": 0.0023, "step": 2395 }, { "epoch": 0.35, "learning_rate": 4.325883201153569e-05, "loss": 0.0031, "step": 2400 }, { "epoch": 0.35, "learning_rate": 4.334895457822639e-05, "loss": 0.0014, "step": 2405 }, { "epoch": 0.35, "learning_rate": 4.343907714491709e-05, "loss": 0.0021, "step": 2410 }, { "epoch": 0.35, "learning_rate": 4.352919971160779e-05, "loss": 0.0021, "step": 2415 }, { "epoch": 0.35, "learning_rate": 4.3619322278298487e-05, "loss": 0.002, "step": 2420 }, { "epoch": 0.35, "learning_rate": 4.370944484498919e-05, "loss": 0.0021, "step": 2425 }, { "epoch": 0.35, "learning_rate": 4.379956741167989e-05, "loss": 0.0025, "step": 2430 }, { "epoch": 0.35, "learning_rate": 4.388968997837059e-05, "loss": 0.0027, "step": 2435 }, { "epoch": 0.35, "learning_rate": 4.3979812545061285e-05, "loss": 0.0019, "step": 2440 }, { "epoch": 0.35, "learning_rate": 4.406993511175199e-05, "loss": 0.0018, "step": 2445 }, { "epoch": 0.35, "learning_rate": 4.416005767844269e-05, "loss": 0.0025, "step": 2450 }, { "epoch": 0.35, "learning_rate": 4.4250180245133385e-05, "loss": 0.0022, "step": 2455 }, { "epoch": 0.35, "learning_rate": 4.434030281182408e-05, "loss": 0.002, "step": 2460 }, { "epoch": 0.36, "learning_rate": 4.443042537851478e-05, "loss": 0.0026, "step": 2465 }, { "epoch": 0.36, "learning_rate": 4.452054794520548e-05, "loss": 0.0021, "step": 2470 }, { "epoch": 0.36, "learning_rate": 4.461067051189618e-05, "loss": 0.002, "step": 2475 }, { "epoch": 0.36, "learning_rate": 4.4700793078586875e-05, "loss": 0.0024, "step": 2480 }, { "epoch": 0.36, "learning_rate": 4.479091564527758e-05, "loss": 0.0024, "step": 2485 }, { "epoch": 0.36, "learning_rate": 4.488103821196828e-05, "loss": 0.0019, "step": 2490 }, { "epoch": 0.36, "learning_rate": 4.4971160778658975e-05, "loss": 0.0023, "step": 2495 }, { "epoch": 0.36, "learning_rate": 4.506128334534967e-05, "loss": 0.0021, "step": 2500 }, { "epoch": 0.36, "learning_rate": 4.515140591204038e-05, "loss": 0.0014, "step": 2505 }, { "epoch": 0.36, "learning_rate": 4.5241528478731076e-05, "loss": 0.0024, "step": 2510 }, { "epoch": 0.36, "learning_rate": 4.5331651045421773e-05, "loss": 0.0026, "step": 2515 }, { "epoch": 0.36, "learning_rate": 4.542177361211247e-05, "loss": 0.0017, "step": 2520 }, { "epoch": 0.36, "learning_rate": 4.5511896178803176e-05, "loss": 0.0017, "step": 2525 }, { "epoch": 0.36, "learning_rate": 4.5602018745493874e-05, "loss": 0.0016, "step": 2530 }, { "epoch": 0.37, "learning_rate": 4.569214131218457e-05, "loss": 0.0025, "step": 2535 }, { "epoch": 0.37, "learning_rate": 4.578226387887527e-05, "loss": 0.0021, "step": 2540 }, { "epoch": 0.37, "learning_rate": 4.5872386445565974e-05, "loss": 0.0018, "step": 2545 }, { "epoch": 0.37, "learning_rate": 4.596250901225667e-05, "loss": 0.0027, "step": 2550 }, { "epoch": 0.37, "learning_rate": 4.605263157894737e-05, "loss": 0.0017, "step": 2555 }, { "epoch": 0.37, "learning_rate": 4.614275414563807e-05, "loss": 0.0023, "step": 2560 }, { "epoch": 0.37, "learning_rate": 4.623287671232877e-05, "loss": 0.0022, "step": 2565 }, { "epoch": 0.37, "learning_rate": 4.632299927901947e-05, "loss": 0.0024, "step": 2570 }, { "epoch": 0.37, "learning_rate": 4.641312184571017e-05, "loss": 0.0026, "step": 2575 }, { "epoch": 0.37, "learning_rate": 4.6503244412400866e-05, "loss": 0.0016, "step": 2580 }, { "epoch": 0.37, "learning_rate": 4.659336697909157e-05, "loss": 0.0026, "step": 2585 }, { "epoch": 0.37, "learning_rate": 4.668348954578227e-05, "loss": 0.0023, "step": 2590 }, { "epoch": 0.37, "learning_rate": 4.677361211247297e-05, "loss": 0.0027, "step": 2595 }, { "epoch": 0.38, "learning_rate": 4.6863734679163665e-05, "loss": 0.0025, "step": 2600 }, { "epoch": 0.38, "learning_rate": 4.695385724585436e-05, "loss": 0.0025, "step": 2605 }, { "epoch": 0.38, "learning_rate": 4.704397981254506e-05, "loss": 0.0022, "step": 2610 }, { "epoch": 0.38, "learning_rate": 4.7134102379235765e-05, "loss": 0.0023, "step": 2615 }, { "epoch": 0.38, "learning_rate": 4.722422494592646e-05, "loss": 0.0032, "step": 2620 }, { "epoch": 0.38, "learning_rate": 4.731434751261716e-05, "loss": 0.0014, "step": 2625 }, { "epoch": 0.38, "learning_rate": 4.740447007930786e-05, "loss": 0.0015, "step": 2630 }, { "epoch": 0.38, "learning_rate": 4.7494592645998557e-05, "loss": 0.0019, "step": 2635 }, { "epoch": 0.38, "learning_rate": 4.7584715212689254e-05, "loss": 0.0025, "step": 2640 }, { "epoch": 0.38, "learning_rate": 4.767483777937996e-05, "loss": 0.0028, "step": 2645 }, { "epoch": 0.38, "learning_rate": 4.776496034607066e-05, "loss": 0.0027, "step": 2650 }, { "epoch": 0.38, "learning_rate": 4.7855082912761355e-05, "loss": 0.0026, "step": 2655 }, { "epoch": 0.38, "learning_rate": 4.794520547945205e-05, "loss": 0.0023, "step": 2660 }, { "epoch": 0.38, "learning_rate": 4.803532804614276e-05, "loss": 0.0026, "step": 2665 }, { "epoch": 0.39, "learning_rate": 4.8125450612833455e-05, "loss": 0.002, "step": 2670 }, { "epoch": 0.39, "learning_rate": 4.821557317952415e-05, "loss": 0.0023, "step": 2675 }, { "epoch": 0.39, "learning_rate": 4.830569574621485e-05, "loss": 0.002, "step": 2680 }, { "epoch": 0.39, "learning_rate": 4.8395818312905556e-05, "loss": 0.0025, "step": 2685 }, { "epoch": 0.39, "learning_rate": 4.8485940879596254e-05, "loss": 0.002, "step": 2690 }, { "epoch": 0.39, "learning_rate": 4.857606344628695e-05, "loss": 0.0023, "step": 2695 }, { "epoch": 0.39, "learning_rate": 4.866618601297765e-05, "loss": 0.0026, "step": 2700 }, { "epoch": 0.39, "learning_rate": 4.8756308579668354e-05, "loss": 0.0025, "step": 2705 }, { "epoch": 0.39, "learning_rate": 4.884643114635905e-05, "loss": 0.002, "step": 2710 }, { "epoch": 0.39, "learning_rate": 4.893655371304975e-05, "loss": 0.0021, "step": 2715 }, { "epoch": 0.39, "learning_rate": 4.902667627974045e-05, "loss": 0.0024, "step": 2720 }, { "epoch": 0.39, "learning_rate": 4.911679884643115e-05, "loss": 0.0017, "step": 2725 }, { "epoch": 0.39, "learning_rate": 4.920692141312185e-05, "loss": 0.0019, "step": 2730 }, { "epoch": 0.39, "learning_rate": 4.929704397981255e-05, "loss": 0.0025, "step": 2735 }, { "epoch": 0.4, "learning_rate": 4.9387166546503246e-05, "loss": 0.0017, "step": 2740 }, { "epoch": 0.4, "learning_rate": 4.947728911319395e-05, "loss": 0.0028, "step": 2745 }, { "epoch": 0.4, "learning_rate": 4.956741167988465e-05, "loss": 0.0024, "step": 2750 }, { "epoch": 0.4, "learning_rate": 4.9657534246575346e-05, "loss": 0.0021, "step": 2755 }, { "epoch": 0.4, "learning_rate": 4.9747656813266044e-05, "loss": 0.0027, "step": 2760 }, { "epoch": 0.4, "learning_rate": 4.983777937995674e-05, "loss": 0.0025, "step": 2765 }, { "epoch": 0.4, "learning_rate": 4.992790194664744e-05, "loss": 0.0022, "step": 2770 }, { "epoch": 0.4, "learning_rate": 4.99979966343457e-05, "loss": 0.0017, "step": 2775 }, { "epoch": 0.4, "learning_rate": 4.998797980607421e-05, "loss": 0.0022, "step": 2780 }, { "epoch": 0.4, "learning_rate": 4.997796297780271e-05, "loss": 0.0023, "step": 2785 }, { "epoch": 0.4, "learning_rate": 4.996794614953121e-05, "loss": 0.002, "step": 2790 }, { "epoch": 0.4, "learning_rate": 4.995792932125972e-05, "loss": 0.0017, "step": 2795 }, { "epoch": 0.4, "learning_rate": 4.994791249298822e-05, "loss": 0.0023, "step": 2800 }, { "epoch": 0.4, "learning_rate": 4.993789566471673e-05, "loss": 0.0024, "step": 2805 }, { "epoch": 0.41, "learning_rate": 4.992787883644523e-05, "loss": 0.0014, "step": 2810 }, { "epoch": 0.41, "learning_rate": 4.9917862008173734e-05, "loss": 0.002, "step": 2815 }, { "epoch": 0.41, "learning_rate": 4.990784517990224e-05, "loss": 0.0018, "step": 2820 }, { "epoch": 0.41, "learning_rate": 4.9897828351630744e-05, "loss": 0.002, "step": 2825 }, { "epoch": 0.41, "learning_rate": 4.9887811523359246e-05, "loss": 0.0026, "step": 2830 }, { "epoch": 0.41, "learning_rate": 4.987779469508775e-05, "loss": 0.0014, "step": 2835 }, { "epoch": 0.41, "learning_rate": 4.986777786681625e-05, "loss": 0.0022, "step": 2840 }, { "epoch": 0.41, "learning_rate": 4.985776103854476e-05, "loss": 0.002, "step": 2845 }, { "epoch": 0.41, "learning_rate": 4.984774421027326e-05, "loss": 0.0024, "step": 2850 }, { "epoch": 0.41, "learning_rate": 4.983772738200176e-05, "loss": 0.0015, "step": 2855 }, { "epoch": 0.41, "learning_rate": 4.9827710553730264e-05, "loss": 0.0024, "step": 2860 }, { "epoch": 0.41, "learning_rate": 4.981769372545877e-05, "loss": 0.0018, "step": 2865 }, { "epoch": 0.41, "learning_rate": 4.9807676897187274e-05, "loss": 0.0023, "step": 2870 }, { "epoch": 0.41, "learning_rate": 4.979766006891578e-05, "loss": 0.002, "step": 2875 }, { "epoch": 0.42, "learning_rate": 4.9787643240644285e-05, "loss": 0.0019, "step": 2880 }, { "epoch": 0.42, "learning_rate": 4.977762641237279e-05, "loss": 0.0026, "step": 2885 }, { "epoch": 0.42, "learning_rate": 4.9767609584101295e-05, "loss": 0.0015, "step": 2890 }, { "epoch": 0.42, "learning_rate": 4.97575927558298e-05, "loss": 0.0023, "step": 2895 }, { "epoch": 0.42, "learning_rate": 4.97475759275583e-05, "loss": 0.0024, "step": 2900 }, { "epoch": 0.42, "learning_rate": 4.973755909928681e-05, "loss": 0.0018, "step": 2905 }, { "epoch": 0.42, "learning_rate": 4.972754227101531e-05, "loss": 0.0018, "step": 2910 }, { "epoch": 0.42, "learning_rate": 4.971752544274381e-05, "loss": 0.0016, "step": 2915 }, { "epoch": 0.42, "learning_rate": 4.970750861447231e-05, "loss": 0.0026, "step": 2920 }, { "epoch": 0.42, "learning_rate": 4.9697491786200815e-05, "loss": 0.0025, "step": 2925 }, { "epoch": 0.42, "learning_rate": 4.9687474957929323e-05, "loss": 0.0024, "step": 2930 }, { "epoch": 0.42, "learning_rate": 4.9677458129657825e-05, "loss": 0.0028, "step": 2935 }, { "epoch": 0.42, "learning_rate": 4.966744130138633e-05, "loss": 0.0021, "step": 2940 }, { "epoch": 0.42, "learning_rate": 4.9657424473114836e-05, "loss": 0.0024, "step": 2945 }, { "epoch": 0.43, "learning_rate": 4.9647407644843344e-05, "loss": 0.0023, "step": 2950 }, { "epoch": 0.43, "learning_rate": 4.9637390816571846e-05, "loss": 0.0023, "step": 2955 }, { "epoch": 0.43, "learning_rate": 4.962737398830035e-05, "loss": 0.0022, "step": 2960 }, { "epoch": 0.43, "learning_rate": 4.961735716002885e-05, "loss": 0.002, "step": 2965 }, { "epoch": 0.43, "learning_rate": 4.960734033175736e-05, "loss": 0.0027, "step": 2970 }, { "epoch": 0.43, "learning_rate": 4.959732350348586e-05, "loss": 0.0015, "step": 2975 }, { "epoch": 0.43, "learning_rate": 4.958730667521436e-05, "loss": 0.0023, "step": 2980 }, { "epoch": 0.43, "learning_rate": 4.9577289846942864e-05, "loss": 0.002, "step": 2985 }, { "epoch": 0.43, "learning_rate": 4.956727301867137e-05, "loss": 0.0019, "step": 2990 }, { "epoch": 0.43, "learning_rate": 4.9557256190399874e-05, "loss": 0.0017, "step": 2995 }, { "epoch": 0.43, "learning_rate": 4.9547239362128376e-05, "loss": 0.002, "step": 3000 }, { "epoch": 0.43, "learning_rate": 4.953722253385688e-05, "loss": 0.0019, "step": 3005 }, { "epoch": 0.43, "learning_rate": 4.952720570558538e-05, "loss": 0.0022, "step": 3010 }, { "epoch": 0.43, "learning_rate": 4.951718887731389e-05, "loss": 0.0017, "step": 3015 }, { "epoch": 0.44, "learning_rate": 4.95071720490424e-05, "loss": 0.0017, "step": 3020 }, { "epoch": 0.44, "learning_rate": 4.94971552207709e-05, "loss": 0.0026, "step": 3025 }, { "epoch": 0.44, "learning_rate": 4.94871383924994e-05, "loss": 0.0019, "step": 3030 }, { "epoch": 0.44, "learning_rate": 4.947712156422791e-05, "loss": 0.0027, "step": 3035 }, { "epoch": 0.44, "learning_rate": 4.946710473595641e-05, "loss": 0.0026, "step": 3040 }, { "epoch": 0.44, "learning_rate": 4.945708790768491e-05, "loss": 0.0019, "step": 3045 }, { "epoch": 0.44, "learning_rate": 4.9447071079413415e-05, "loss": 0.0018, "step": 3050 }, { "epoch": 0.44, "learning_rate": 4.9437054251141924e-05, "loss": 0.002, "step": 3055 }, { "epoch": 0.44, "learning_rate": 4.9427037422870425e-05, "loss": 0.0025, "step": 3060 }, { "epoch": 0.44, "learning_rate": 4.941702059459893e-05, "loss": 0.0022, "step": 3065 }, { "epoch": 0.44, "learning_rate": 4.940700376632743e-05, "loss": 0.0016, "step": 3070 }, { "epoch": 0.44, "learning_rate": 4.939698693805593e-05, "loss": 0.0024, "step": 3075 }, { "epoch": 0.44, "learning_rate": 4.938697010978444e-05, "loss": 0.0015, "step": 3080 }, { "epoch": 0.44, "learning_rate": 4.937695328151294e-05, "loss": 0.0018, "step": 3085 }, { "epoch": 0.45, "learning_rate": 4.936693645324145e-05, "loss": 0.0025, "step": 3090 }, { "epoch": 0.45, "learning_rate": 4.935691962496995e-05, "loss": 0.0017, "step": 3095 }, { "epoch": 0.45, "learning_rate": 4.934690279669846e-05, "loss": 0.0019, "step": 3100 }, { "epoch": 0.45, "learning_rate": 4.933688596842696e-05, "loss": 0.0018, "step": 3105 }, { "epoch": 0.45, "learning_rate": 4.9326869140155464e-05, "loss": 0.0019, "step": 3110 }, { "epoch": 0.45, "learning_rate": 4.9316852311883966e-05, "loss": 0.0022, "step": 3115 }, { "epoch": 0.45, "learning_rate": 4.9306835483612475e-05, "loss": 0.0023, "step": 3120 }, { "epoch": 0.45, "learning_rate": 4.9296818655340976e-05, "loss": 0.0025, "step": 3125 }, { "epoch": 0.45, "learning_rate": 4.928680182706948e-05, "loss": 0.0023, "step": 3130 }, { "epoch": 0.45, "learning_rate": 4.927678499879798e-05, "loss": 0.0024, "step": 3135 }, { "epoch": 0.45, "learning_rate": 4.926676817052649e-05, "loss": 0.0023, "step": 3140 }, { "epoch": 0.45, "learning_rate": 4.925675134225499e-05, "loss": 0.0018, "step": 3145 }, { "epoch": 0.45, "learning_rate": 4.924673451398349e-05, "loss": 0.0022, "step": 3150 }, { "epoch": 0.46, "learning_rate": 4.9236717685711994e-05, "loss": 0.0023, "step": 3155 }, { "epoch": 0.46, "learning_rate": 4.92267008574405e-05, "loss": 0.0017, "step": 3160 }, { "epoch": 0.46, "learning_rate": 4.9216684029169005e-05, "loss": 0.0022, "step": 3165 }, { "epoch": 0.46, "learning_rate": 4.920666720089751e-05, "loss": 0.0028, "step": 3170 }, { "epoch": 0.46, "learning_rate": 4.9196650372626015e-05, "loss": 0.0022, "step": 3175 }, { "epoch": 0.46, "learning_rate": 4.918663354435452e-05, "loss": 0.0021, "step": 3180 }, { "epoch": 0.46, "learning_rate": 4.9176616716083026e-05, "loss": 0.0016, "step": 3185 }, { "epoch": 0.46, "learning_rate": 4.916659988781153e-05, "loss": 0.0019, "step": 3190 }, { "epoch": 0.46, "learning_rate": 4.915658305954003e-05, "loss": 0.0017, "step": 3195 }, { "epoch": 0.46, "learning_rate": 4.914656623126853e-05, "loss": 0.0023, "step": 3200 }, { "epoch": 0.46, "learning_rate": 4.913654940299704e-05, "loss": 0.0017, "step": 3205 }, { "epoch": 0.46, "learning_rate": 4.912653257472554e-05, "loss": 0.002, "step": 3210 }, { "epoch": 0.46, "learning_rate": 4.9116515746454043e-05, "loss": 0.0013, "step": 3215 }, { "epoch": 0.46, "learning_rate": 4.9106498918182545e-05, "loss": 0.0024, "step": 3220 }, { "epoch": 0.47, "learning_rate": 4.9096482089911054e-05, "loss": 0.0017, "step": 3225 }, { "epoch": 0.47, "learning_rate": 4.9086465261639556e-05, "loss": 0.0023, "step": 3230 }, { "epoch": 0.47, "learning_rate": 4.907644843336806e-05, "loss": 0.002, "step": 3235 }, { "epoch": 0.47, "learning_rate": 4.9066431605096566e-05, "loss": 0.0016, "step": 3240 }, { "epoch": 0.47, "learning_rate": 4.905641477682507e-05, "loss": 0.0017, "step": 3245 }, { "epoch": 0.47, "learning_rate": 4.904639794855358e-05, "loss": 0.002, "step": 3250 }, { "epoch": 0.47, "learning_rate": 4.903638112028208e-05, "loss": 0.002, "step": 3255 }, { "epoch": 0.47, "learning_rate": 4.902636429201058e-05, "loss": 0.0026, "step": 3260 }, { "epoch": 0.47, "learning_rate": 4.901634746373908e-05, "loss": 0.002, "step": 3265 }, { "epoch": 0.47, "learning_rate": 4.900633063546759e-05, "loss": 0.0023, "step": 3270 }, { "epoch": 0.47, "learning_rate": 4.899631380719609e-05, "loss": 0.002, "step": 3275 }, { "epoch": 0.47, "learning_rate": 4.8986296978924594e-05, "loss": 0.0019, "step": 3280 }, { "epoch": 0.47, "learning_rate": 4.8976280150653096e-05, "loss": 0.0017, "step": 3285 }, { "epoch": 0.47, "learning_rate": 4.8966263322381605e-05, "loss": 0.0017, "step": 3290 }, { "epoch": 0.48, "learning_rate": 4.895624649411011e-05, "loss": 0.0022, "step": 3295 }, { "epoch": 0.48, "learning_rate": 4.894622966583861e-05, "loss": 0.0015, "step": 3300 }, { "epoch": 0.48, "learning_rate": 4.893621283756711e-05, "loss": 0.0019, "step": 3305 }, { "epoch": 0.48, "learning_rate": 4.892619600929562e-05, "loss": 0.0021, "step": 3310 }, { "epoch": 0.48, "learning_rate": 4.891617918102413e-05, "loss": 0.002, "step": 3315 }, { "epoch": 0.48, "learning_rate": 4.890616235275263e-05, "loss": 0.0023, "step": 3320 }, { "epoch": 0.48, "learning_rate": 4.889614552448113e-05, "loss": 0.0017, "step": 3325 }, { "epoch": 0.48, "learning_rate": 4.888612869620963e-05, "loss": 0.0022, "step": 3330 }, { "epoch": 0.48, "learning_rate": 4.887611186793814e-05, "loss": 0.0021, "step": 3335 }, { "epoch": 0.48, "learning_rate": 4.8866095039666644e-05, "loss": 0.0025, "step": 3340 }, { "epoch": 0.48, "learning_rate": 4.8856078211395145e-05, "loss": 0.0018, "step": 3345 }, { "epoch": 0.48, "learning_rate": 4.884606138312365e-05, "loss": 0.0022, "step": 3350 }, { "epoch": 0.48, "learning_rate": 4.8836044554852156e-05, "loss": 0.0017, "step": 3355 }, { "epoch": 0.48, "learning_rate": 4.882602772658066e-05, "loss": 0.0016, "step": 3360 }, { "epoch": 0.49, "learning_rate": 4.881601089830916e-05, "loss": 0.0021, "step": 3365 }, { "epoch": 0.49, "learning_rate": 4.880599407003766e-05, "loss": 0.0021, "step": 3370 }, { "epoch": 0.49, "learning_rate": 4.879597724176617e-05, "loss": 0.0021, "step": 3375 }, { "epoch": 0.49, "learning_rate": 4.878596041349467e-05, "loss": 0.0015, "step": 3380 }, { "epoch": 0.49, "learning_rate": 4.877594358522318e-05, "loss": 0.0014, "step": 3385 }, { "epoch": 0.49, "learning_rate": 4.876592675695168e-05, "loss": 0.0025, "step": 3390 }, { "epoch": 0.49, "learning_rate": 4.8755909928680184e-05, "loss": 0.0025, "step": 3395 }, { "epoch": 0.49, "learning_rate": 4.874589310040869e-05, "loss": 0.0023, "step": 3400 }, { "epoch": 0.49, "learning_rate": 4.8735876272137195e-05, "loss": 0.002, "step": 3405 }, { "epoch": 0.49, "learning_rate": 4.8725859443865696e-05, "loss": 0.0025, "step": 3410 }, { "epoch": 0.49, "learning_rate": 4.87158426155942e-05, "loss": 0.0011, "step": 3415 }, { "epoch": 0.49, "learning_rate": 4.870582578732271e-05, "loss": 0.002, "step": 3420 }, { "epoch": 0.49, "learning_rate": 4.869580895905121e-05, "loss": 0.0019, "step": 3425 }, { "epoch": 0.49, "learning_rate": 4.868579213077971e-05, "loss": 0.0019, "step": 3430 }, { "epoch": 0.5, "learning_rate": 4.867577530250821e-05, "loss": 0.0014, "step": 3435 }, { "epoch": 0.5, "learning_rate": 4.866575847423672e-05, "loss": 0.0022, "step": 3440 }, { "epoch": 0.5, "learning_rate": 4.865574164596522e-05, "loss": 0.0018, "step": 3445 }, { "epoch": 0.5, "learning_rate": 4.8645724817693725e-05, "loss": 0.0018, "step": 3450 }, { "epoch": 0.5, "learning_rate": 4.863570798942223e-05, "loss": 0.0018, "step": 3455 }, { "epoch": 0.5, "learning_rate": 4.8625691161150735e-05, "loss": 0.0017, "step": 3460 }, { "epoch": 0.5, "learning_rate": 4.8615674332879244e-05, "loss": 0.0021, "step": 3465 }, { "epoch": 0.5, "learning_rate": 4.8605657504607746e-05, "loss": 0.0017, "step": 3470 }, { "epoch": 0.5, "learning_rate": 4.859564067633625e-05, "loss": 0.0021, "step": 3475 }, { "epoch": 0.5, "learning_rate": 4.858562384806475e-05, "loss": 0.002, "step": 3480 }, { "epoch": 0.5, "learning_rate": 4.857560701979326e-05, "loss": 0.0016, "step": 3485 }, { "epoch": 0.5, "learning_rate": 4.856559019152176e-05, "loss": 0.0019, "step": 3490 }, { "epoch": 0.5, "learning_rate": 4.855557336325026e-05, "loss": 0.0016, "step": 3495 }, { "epoch": 0.5, "learning_rate": 4.8545556534978763e-05, "loss": 0.0017, "step": 3500 }, { "epoch": 0.51, "learning_rate": 4.853553970670727e-05, "loss": 0.0016, "step": 3505 }, { "epoch": 0.51, "learning_rate": 4.8525522878435774e-05, "loss": 0.0026, "step": 3510 }, { "epoch": 0.51, "learning_rate": 4.8515506050164276e-05, "loss": 0.0015, "step": 3515 }, { "epoch": 0.51, "learning_rate": 4.850548922189278e-05, "loss": 0.0022, "step": 3520 }, { "epoch": 0.51, "learning_rate": 4.8495472393621286e-05, "loss": 0.0015, "step": 3525 }, { "epoch": 0.51, "learning_rate": 4.848545556534979e-05, "loss": 0.0017, "step": 3530 }, { "epoch": 0.51, "learning_rate": 4.84754387370783e-05, "loss": 0.0017, "step": 3535 }, { "epoch": 0.51, "learning_rate": 4.84654219088068e-05, "loss": 0.002, "step": 3540 }, { "epoch": 0.51, "learning_rate": 4.845540508053531e-05, "loss": 0.0017, "step": 3545 }, { "epoch": 0.51, "learning_rate": 4.844538825226381e-05, "loss": 0.0016, "step": 3550 }, { "epoch": 0.51, "learning_rate": 4.843537142399231e-05, "loss": 0.0019, "step": 3555 }, { "epoch": 0.51, "learning_rate": 4.842535459572081e-05, "loss": 0.0016, "step": 3560 }, { "epoch": 0.51, "learning_rate": 4.8415337767449314e-05, "loss": 0.0016, "step": 3565 }, { "epoch": 0.51, "learning_rate": 4.840532093917782e-05, "loss": 0.002, "step": 3570 }, { "epoch": 0.52, "learning_rate": 4.8395304110906325e-05, "loss": 0.0012, "step": 3575 }, { "epoch": 0.52, "learning_rate": 4.838528728263483e-05, "loss": 0.0018, "step": 3580 }, { "epoch": 0.52, "learning_rate": 4.837527045436333e-05, "loss": 0.0022, "step": 3585 }, { "epoch": 0.52, "learning_rate": 4.836525362609184e-05, "loss": 0.0022, "step": 3590 }, { "epoch": 0.52, "learning_rate": 4.835523679782034e-05, "loss": 0.0019, "step": 3595 }, { "epoch": 0.52, "learning_rate": 4.834521996954884e-05, "loss": 0.0016, "step": 3600 }, { "epoch": 0.52, "learning_rate": 4.833520314127735e-05, "loss": 0.0017, "step": 3605 }, { "epoch": 0.52, "learning_rate": 4.832518631300585e-05, "loss": 0.0016, "step": 3610 }, { "epoch": 0.52, "learning_rate": 4.831516948473436e-05, "loss": 0.0022, "step": 3615 }, { "epoch": 0.52, "learning_rate": 4.830515265646286e-05, "loss": 0.0019, "step": 3620 }, { "epoch": 0.52, "learning_rate": 4.8295135828191364e-05, "loss": 0.0021, "step": 3625 }, { "epoch": 0.52, "learning_rate": 4.8285118999919865e-05, "loss": 0.002, "step": 3630 }, { "epoch": 0.52, "learning_rate": 4.8275102171648374e-05, "loss": 0.0021, "step": 3635 }, { "epoch": 0.53, "learning_rate": 4.8265085343376876e-05, "loss": 0.002, "step": 3640 }, { "epoch": 0.53, "learning_rate": 4.825506851510538e-05, "loss": 0.0028, "step": 3645 }, { "epoch": 0.53, "learning_rate": 4.824505168683388e-05, "loss": 0.002, "step": 3650 }, { "epoch": 0.53, "learning_rate": 4.823503485856239e-05, "loss": 0.0016, "step": 3655 }, { "epoch": 0.53, "learning_rate": 4.822501803029089e-05, "loss": 0.0011, "step": 3660 }, { "epoch": 0.53, "learning_rate": 4.821500120201939e-05, "loss": 0.0023, "step": 3665 }, { "epoch": 0.53, "learning_rate": 4.8204984373747894e-05, "loss": 0.0022, "step": 3670 }, { "epoch": 0.53, "learning_rate": 4.81949675454764e-05, "loss": 0.0017, "step": 3675 }, { "epoch": 0.53, "learning_rate": 4.8184950717204904e-05, "loss": 0.0013, "step": 3680 }, { "epoch": 0.53, "learning_rate": 4.817493388893341e-05, "loss": 0.0023, "step": 3685 }, { "epoch": 0.53, "learning_rate": 4.8164917060661915e-05, "loss": 0.002, "step": 3690 }, { "epoch": 0.53, "learning_rate": 4.815490023239042e-05, "loss": 0.0018, "step": 3695 }, { "epoch": 0.53, "learning_rate": 4.8144883404118925e-05, "loss": 0.0026, "step": 3700 }, { "epoch": 0.53, "learning_rate": 4.813486657584743e-05, "loss": 0.0014, "step": 3705 }, { "epoch": 0.54, "learning_rate": 4.812484974757593e-05, "loss": 0.0016, "step": 3710 }, { "epoch": 0.54, "learning_rate": 4.811483291930443e-05, "loss": 0.0015, "step": 3715 }, { "epoch": 0.54, "learning_rate": 4.810481609103294e-05, "loss": 0.002, "step": 3720 }, { "epoch": 0.54, "learning_rate": 4.809479926276144e-05, "loss": 0.0021, "step": 3725 }, { "epoch": 0.54, "learning_rate": 4.808478243448994e-05, "loss": 0.0026, "step": 3730 }, { "epoch": 0.54, "learning_rate": 4.8074765606218445e-05, "loss": 0.0014, "step": 3735 }, { "epoch": 0.54, "learning_rate": 4.806474877794695e-05, "loss": 0.0014, "step": 3740 }, { "epoch": 0.54, "learning_rate": 4.8054731949675455e-05, "loss": 0.0015, "step": 3745 }, { "epoch": 0.54, "learning_rate": 4.804471512140396e-05, "loss": 0.0025, "step": 3750 }, { "epoch": 0.54, "learning_rate": 4.8034698293132466e-05, "loss": 0.0018, "step": 3755 }, { "epoch": 0.54, "learning_rate": 4.8024681464860974e-05, "loss": 0.002, "step": 3760 }, { "epoch": 0.54, "learning_rate": 4.8014664636589476e-05, "loss": 0.0017, "step": 3765 }, { "epoch": 0.54, "learning_rate": 4.800464780831798e-05, "loss": 0.0024, "step": 3770 }, { "epoch": 0.54, "learning_rate": 4.799463098004648e-05, "loss": 0.0017, "step": 3775 }, { "epoch": 0.55, "learning_rate": 4.798461415177499e-05, "loss": 0.0022, "step": 3780 }, { "epoch": 0.55, "learning_rate": 4.797459732350349e-05, "loss": 0.0016, "step": 3785 }, { "epoch": 0.55, "learning_rate": 4.796458049523199e-05, "loss": 0.0021, "step": 3790 }, { "epoch": 0.55, "learning_rate": 4.7954563666960494e-05, "loss": 0.0019, "step": 3795 }, { "epoch": 0.55, "learning_rate": 4.7944546838688996e-05, "loss": 0.0015, "step": 3800 }, { "epoch": 0.55, "learning_rate": 4.7934530010417504e-05, "loss": 0.0017, "step": 3805 }, { "epoch": 0.55, "learning_rate": 4.7924513182146006e-05, "loss": 0.0022, "step": 3810 }, { "epoch": 0.55, "learning_rate": 4.791449635387451e-05, "loss": 0.0025, "step": 3815 }, { "epoch": 0.55, "learning_rate": 4.790447952560301e-05, "loss": 0.0017, "step": 3820 }, { "epoch": 0.55, "learning_rate": 4.789446269733152e-05, "loss": 0.0021, "step": 3825 }, { "epoch": 0.55, "learning_rate": 4.788444586906003e-05, "loss": 0.002, "step": 3830 }, { "epoch": 0.55, "learning_rate": 4.787442904078853e-05, "loss": 0.0022, "step": 3835 }, { "epoch": 0.55, "learning_rate": 4.786441221251703e-05, "loss": 0.0017, "step": 3840 }, { "epoch": 0.55, "learning_rate": 4.785439538424554e-05, "loss": 0.0019, "step": 3845 }, { "epoch": 0.56, "learning_rate": 4.784437855597404e-05, "loss": 0.0019, "step": 3850 }, { "epoch": 0.56, "learning_rate": 4.783436172770254e-05, "loss": 0.0015, "step": 3855 }, { "epoch": 0.56, "learning_rate": 4.7824344899431045e-05, "loss": 0.0022, "step": 3860 }, { "epoch": 0.56, "learning_rate": 4.781432807115955e-05, "loss": 0.0019, "step": 3865 }, { "epoch": 0.56, "learning_rate": 4.7804311242888055e-05, "loss": 0.0016, "step": 3870 }, { "epoch": 0.56, "learning_rate": 4.779429441461656e-05, "loss": 0.0016, "step": 3875 }, { "epoch": 0.56, "learning_rate": 4.778427758634506e-05, "loss": 0.0016, "step": 3880 }, { "epoch": 0.56, "learning_rate": 4.777426075807356e-05, "loss": 0.0024, "step": 3885 }, { "epoch": 0.56, "learning_rate": 4.776424392980207e-05, "loss": 0.0023, "step": 3890 }, { "epoch": 0.56, "learning_rate": 4.775422710153057e-05, "loss": 0.0018, "step": 3895 }, { "epoch": 0.56, "learning_rate": 4.774421027325908e-05, "loss": 0.0019, "step": 3900 }, { "epoch": 0.56, "learning_rate": 4.773419344498758e-05, "loss": 0.0017, "step": 3905 }, { "epoch": 0.56, "learning_rate": 4.772417661671609e-05, "loss": 0.0021, "step": 3910 }, { "epoch": 0.56, "learning_rate": 4.771415978844459e-05, "loss": 0.0016, "step": 3915 }, { "epoch": 0.57, "learning_rate": 4.7704142960173094e-05, "loss": 0.0024, "step": 3920 }, { "epoch": 0.57, "learning_rate": 4.7694126131901596e-05, "loss": 0.0023, "step": 3925 }, { "epoch": 0.57, "learning_rate": 4.7684109303630105e-05, "loss": 0.0017, "step": 3930 }, { "epoch": 0.57, "learning_rate": 4.7674092475358606e-05, "loss": 0.0016, "step": 3935 }, { "epoch": 0.57, "learning_rate": 4.766407564708711e-05, "loss": 0.0021, "step": 3940 }, { "epoch": 0.57, "learning_rate": 4.765405881881561e-05, "loss": 0.0021, "step": 3945 }, { "epoch": 0.57, "learning_rate": 4.764404199054411e-05, "loss": 0.002, "step": 3950 }, { "epoch": 0.57, "learning_rate": 4.763402516227262e-05, "loss": 0.0016, "step": 3955 }, { "epoch": 0.57, "learning_rate": 4.762400833400112e-05, "loss": 0.0023, "step": 3960 }, { "epoch": 0.57, "learning_rate": 4.7613991505729624e-05, "loss": 0.0022, "step": 3965 }, { "epoch": 0.57, "learning_rate": 4.760397467745813e-05, "loss": 0.0022, "step": 3970 }, { "epoch": 0.57, "learning_rate": 4.7593957849186635e-05, "loss": 0.0022, "step": 3975 }, { "epoch": 0.57, "learning_rate": 4.758394102091514e-05, "loss": 0.0017, "step": 3980 }, { "epoch": 0.57, "learning_rate": 4.7573924192643645e-05, "loss": 0.0017, "step": 3985 }, { "epoch": 0.58, "learning_rate": 4.756390736437215e-05, "loss": 0.0017, "step": 3990 }, { "epoch": 0.58, "learning_rate": 4.7553890536100656e-05, "loss": 0.0015, "step": 3995 }, { "epoch": 0.58, "learning_rate": 4.754387370782916e-05, "loss": 0.0017, "step": 4000 }, { "epoch": 0.58, "learning_rate": 4.753385687955766e-05, "loss": 0.002, "step": 4005 }, { "epoch": 0.58, "learning_rate": 4.752384005128616e-05, "loss": 0.0015, "step": 4010 }, { "epoch": 0.58, "learning_rate": 4.751382322301467e-05, "loss": 0.0022, "step": 4015 }, { "epoch": 0.58, "learning_rate": 4.750380639474317e-05, "loss": 0.0014, "step": 4020 }, { "epoch": 0.58, "learning_rate": 4.749378956647167e-05, "loss": 0.002, "step": 4025 }, { "epoch": 0.58, "learning_rate": 4.7483772738200175e-05, "loss": 0.0018, "step": 4030 }, { "epoch": 0.58, "learning_rate": 4.747375590992868e-05, "loss": 0.0018, "step": 4035 }, { "epoch": 0.58, "learning_rate": 4.7463739081657186e-05, "loss": 0.0016, "step": 4040 }, { "epoch": 0.58, "learning_rate": 4.745372225338569e-05, "loss": 0.0016, "step": 4045 }, { "epoch": 0.58, "learning_rate": 4.7443705425114196e-05, "loss": 0.0018, "step": 4050 }, { "epoch": 0.58, "learning_rate": 4.74336885968427e-05, "loss": 0.0018, "step": 4055 }, { "epoch": 0.59, "learning_rate": 4.7423671768571207e-05, "loss": 0.0017, "step": 4060 }, { "epoch": 0.59, "learning_rate": 4.741365494029971e-05, "loss": 0.0021, "step": 4065 }, { "epoch": 0.59, "learning_rate": 4.740363811202821e-05, "loss": 0.0016, "step": 4070 }, { "epoch": 0.59, "learning_rate": 4.739362128375671e-05, "loss": 0.0015, "step": 4075 }, { "epoch": 0.59, "learning_rate": 4.738360445548522e-05, "loss": 0.0016, "step": 4080 }, { "epoch": 0.59, "learning_rate": 4.737358762721372e-05, "loss": 0.0018, "step": 4085 }, { "epoch": 0.59, "learning_rate": 4.7363570798942224e-05, "loss": 0.0022, "step": 4090 }, { "epoch": 0.59, "learning_rate": 4.7353553970670726e-05, "loss": 0.0018, "step": 4095 }, { "epoch": 0.59, "learning_rate": 4.734353714239923e-05, "loss": 0.0017, "step": 4100 }, { "epoch": 0.59, "learning_rate": 4.733352031412774e-05, "loss": 0.0014, "step": 4105 }, { "epoch": 0.59, "learning_rate": 4.732350348585624e-05, "loss": 0.0016, "step": 4110 }, { "epoch": 0.59, "learning_rate": 4.731348665758474e-05, "loss": 0.0015, "step": 4115 }, { "epoch": 0.59, "learning_rate": 4.730346982931325e-05, "loss": 0.0017, "step": 4120 }, { "epoch": 0.59, "learning_rate": 4.729345300104176e-05, "loss": 0.0027, "step": 4125 }, { "epoch": 0.6, "learning_rate": 4.728343617277026e-05, "loss": 0.0017, "step": 4130 }, { "epoch": 0.6, "learning_rate": 4.727341934449876e-05, "loss": 0.0015, "step": 4135 }, { "epoch": 0.6, "learning_rate": 4.726340251622726e-05, "loss": 0.0016, "step": 4140 }, { "epoch": 0.6, "learning_rate": 4.725338568795577e-05, "loss": 0.0013, "step": 4145 }, { "epoch": 0.6, "learning_rate": 4.7243368859684274e-05, "loss": 0.0014, "step": 4150 }, { "epoch": 0.6, "learning_rate": 4.7233352031412775e-05, "loss": 0.002, "step": 4155 }, { "epoch": 0.6, "learning_rate": 4.722333520314128e-05, "loss": 0.0014, "step": 4160 }, { "epoch": 0.6, "learning_rate": 4.7213318374869786e-05, "loss": 0.0017, "step": 4165 }, { "epoch": 0.6, "learning_rate": 4.720330154659829e-05, "loss": 0.0015, "step": 4170 }, { "epoch": 0.6, "learning_rate": 4.719328471832679e-05, "loss": 0.0019, "step": 4175 }, { "epoch": 0.6, "learning_rate": 4.718326789005529e-05, "loss": 0.0015, "step": 4180 }, { "epoch": 0.6, "learning_rate": 4.717325106178379e-05, "loss": 0.0016, "step": 4185 }, { "epoch": 0.6, "learning_rate": 4.71632342335123e-05, "loss": 0.0018, "step": 4190 }, { "epoch": 0.61, "learning_rate": 4.715321740524081e-05, "loss": 0.0022, "step": 4195 }, { "epoch": 0.61, "learning_rate": 4.714320057696931e-05, "loss": 0.0014, "step": 4200 }, { "epoch": 0.61, "learning_rate": 4.7133183748697814e-05, "loss": 0.0019, "step": 4205 }, { "epoch": 0.61, "learning_rate": 4.712316692042632e-05, "loss": 0.0016, "step": 4210 }, { "epoch": 0.61, "learning_rate": 4.7113150092154825e-05, "loss": 0.0018, "step": 4215 }, { "epoch": 0.61, "learning_rate": 4.7103133263883326e-05, "loss": 0.0013, "step": 4220 }, { "epoch": 0.61, "learning_rate": 4.709311643561183e-05, "loss": 0.0018, "step": 4225 }, { "epoch": 0.61, "learning_rate": 4.708309960734034e-05, "loss": 0.0013, "step": 4230 }, { "epoch": 0.61, "learning_rate": 4.707308277906884e-05, "loss": 0.0017, "step": 4235 }, { "epoch": 0.61, "learning_rate": 4.706306595079734e-05, "loss": 0.0016, "step": 4240 }, { "epoch": 0.61, "learning_rate": 4.705304912252584e-05, "loss": 0.0018, "step": 4245 }, { "epoch": 0.61, "learning_rate": 4.704303229425435e-05, "loss": 0.0021, "step": 4250 }, { "epoch": 0.61, "learning_rate": 4.703301546598285e-05, "loss": 0.0014, "step": 4255 }, { "epoch": 0.61, "learning_rate": 4.7022998637711355e-05, "loss": 0.0018, "step": 4260 }, { "epoch": 0.62, "learning_rate": 4.701298180943986e-05, "loss": 0.0014, "step": 4265 }, { "epoch": 0.62, "learning_rate": 4.7002964981168365e-05, "loss": 0.0018, "step": 4270 }, { "epoch": 0.62, "learning_rate": 4.6992948152896874e-05, "loss": 0.0021, "step": 4275 }, { "epoch": 0.62, "learning_rate": 4.6982931324625376e-05, "loss": 0.0017, "step": 4280 }, { "epoch": 0.62, "learning_rate": 4.697291449635388e-05, "loss": 0.0016, "step": 4285 }, { "epoch": 0.62, "learning_rate": 4.696289766808238e-05, "loss": 0.0011, "step": 4290 }, { "epoch": 0.62, "learning_rate": 4.695288083981089e-05, "loss": 0.0014, "step": 4295 }, { "epoch": 0.62, "learning_rate": 4.694286401153939e-05, "loss": 0.0015, "step": 4300 }, { "epoch": 0.62, "learning_rate": 4.693284718326789e-05, "loss": 0.0017, "step": 4305 }, { "epoch": 0.62, "learning_rate": 4.692283035499639e-05, "loss": 0.0016, "step": 4310 }, { "epoch": 0.62, "learning_rate": 4.69128135267249e-05, "loss": 0.0014, "step": 4315 }, { "epoch": 0.62, "learning_rate": 4.6902796698453404e-05, "loss": 0.0013, "step": 4320 }, { "epoch": 0.62, "learning_rate": 4.6892779870181906e-05, "loss": 0.0013, "step": 4325 }, { "epoch": 0.62, "learning_rate": 4.688276304191041e-05, "loss": 0.0022, "step": 4330 }, { "epoch": 0.63, "learning_rate": 4.6872746213638916e-05, "loss": 0.002, "step": 4335 }, { "epoch": 0.63, "learning_rate": 4.686272938536742e-05, "loss": 0.0013, "step": 4340 }, { "epoch": 0.63, "learning_rate": 4.6852712557095927e-05, "loss": 0.0017, "step": 4345 }, { "epoch": 0.63, "learning_rate": 4.684269572882443e-05, "loss": 0.0013, "step": 4350 }, { "epoch": 0.63, "learning_rate": 4.683267890055293e-05, "loss": 0.0022, "step": 4355 }, { "epoch": 0.63, "learning_rate": 4.682266207228144e-05, "loss": 0.0016, "step": 4360 }, { "epoch": 0.63, "learning_rate": 4.681264524400994e-05, "loss": 0.0017, "step": 4365 }, { "epoch": 0.63, "learning_rate": 4.680262841573844e-05, "loss": 0.0013, "step": 4370 }, { "epoch": 0.63, "learning_rate": 4.6792611587466944e-05, "loss": 0.0015, "step": 4375 }, { "epoch": 0.63, "learning_rate": 4.678259475919545e-05, "loss": 0.0012, "step": 4380 }, { "epoch": 0.63, "learning_rate": 4.6772577930923955e-05, "loss": 0.0012, "step": 4385 }, { "epoch": 0.63, "learning_rate": 4.676256110265246e-05, "loss": 0.002, "step": 4390 }, { "epoch": 0.63, "learning_rate": 4.675254427438096e-05, "loss": 0.0009, "step": 4395 }, { "epoch": 0.63, "learning_rate": 4.674252744610947e-05, "loss": 0.0016, "step": 4400 }, { "epoch": 0.64, "learning_rate": 4.673251061783797e-05, "loss": 0.0018, "step": 4405 }, { "epoch": 0.64, "learning_rate": 4.672249378956647e-05, "loss": 0.0017, "step": 4410 }, { "epoch": 0.64, "learning_rate": 4.671247696129498e-05, "loss": 0.0017, "step": 4415 }, { "epoch": 0.64, "learning_rate": 4.670246013302348e-05, "loss": 0.0018, "step": 4420 }, { "epoch": 0.64, "learning_rate": 4.669244330475199e-05, "loss": 0.0016, "step": 4425 }, { "epoch": 0.64, "learning_rate": 4.668242647648049e-05, "loss": 0.0014, "step": 4430 }, { "epoch": 0.64, "learning_rate": 4.6672409648208994e-05, "loss": 0.0022, "step": 4435 }, { "epoch": 0.64, "learning_rate": 4.6662392819937495e-05, "loss": 0.0015, "step": 4440 }, { "epoch": 0.64, "learning_rate": 4.6652375991666004e-05, "loss": 0.0014, "step": 4445 }, { "epoch": 0.64, "learning_rate": 4.6642359163394506e-05, "loss": 0.0014, "step": 4450 }, { "epoch": 0.64, "learning_rate": 4.663234233512301e-05, "loss": 0.0017, "step": 4455 }, { "epoch": 0.64, "learning_rate": 4.662232550685151e-05, "loss": 0.0014, "step": 4460 }, { "epoch": 0.64, "learning_rate": 4.661230867858002e-05, "loss": 0.0011, "step": 4465 }, { "epoch": 0.64, "learning_rate": 4.660229185030852e-05, "loss": 0.0013, "step": 4470 }, { "epoch": 0.65, "learning_rate": 4.659227502203702e-05, "loss": 0.001, "step": 4475 }, { "epoch": 0.65, "learning_rate": 4.6582258193765524e-05, "loss": 0.0019, "step": 4480 }, { "epoch": 0.65, "learning_rate": 4.657224136549403e-05, "loss": 0.0018, "step": 4485 }, { "epoch": 0.65, "learning_rate": 4.6562224537222534e-05, "loss": 0.0017, "step": 4490 }, { "epoch": 0.65, "learning_rate": 4.655220770895104e-05, "loss": 0.0017, "step": 4495 }, { "epoch": 0.65, "learning_rate": 4.6542190880679545e-05, "loss": 0.0016, "step": 4500 }, { "epoch": 0.65, "learning_rate": 4.6532174052408046e-05, "loss": 0.0015, "step": 4505 }, { "epoch": 0.65, "learning_rate": 4.6522157224136555e-05, "loss": 0.0019, "step": 4510 }, { "epoch": 0.65, "learning_rate": 4.651214039586506e-05, "loss": 0.0017, "step": 4515 }, { "epoch": 0.65, "learning_rate": 4.650212356759356e-05, "loss": 0.0016, "step": 4520 }, { "epoch": 0.65, "learning_rate": 4.649210673932206e-05, "loss": 0.0013, "step": 4525 }, { "epoch": 0.65, "learning_rate": 4.648208991105057e-05, "loss": 0.002, "step": 4530 }, { "epoch": 0.65, "learning_rate": 4.647207308277907e-05, "loss": 0.0018, "step": 4535 }, { "epoch": 0.65, "learning_rate": 4.646205625450757e-05, "loss": 0.0008, "step": 4540 }, { "epoch": 0.66, "learning_rate": 4.6452039426236075e-05, "loss": 0.0015, "step": 4545 }, { "epoch": 0.66, "learning_rate": 4.644202259796458e-05, "loss": 0.0013, "step": 4550 }, { "epoch": 0.66, "learning_rate": 4.6432005769693085e-05, "loss": 0.0019, "step": 4555 }, { "epoch": 0.66, "learning_rate": 4.642198894142159e-05, "loss": 0.0016, "step": 4560 }, { "epoch": 0.66, "learning_rate": 4.6411972113150096e-05, "loss": 0.0012, "step": 4565 }, { "epoch": 0.66, "learning_rate": 4.6401955284878604e-05, "loss": 0.0015, "step": 4570 }, { "epoch": 0.66, "learning_rate": 4.6391938456607106e-05, "loss": 0.0018, "step": 4575 }, { "epoch": 0.66, "learning_rate": 4.638192162833561e-05, "loss": 0.0016, "step": 4580 }, { "epoch": 0.66, "learning_rate": 4.637190480006411e-05, "loss": 0.0019, "step": 4585 }, { "epoch": 0.66, "learning_rate": 4.636188797179261e-05, "loss": 0.0011, "step": 4590 }, { "epoch": 0.66, "learning_rate": 4.635187114352112e-05, "loss": 0.0015, "step": 4595 }, { "epoch": 0.66, "learning_rate": 4.634185431524962e-05, "loss": 0.0009, "step": 4600 }, { "epoch": 0.66, "learning_rate": 4.6331837486978124e-05, "loss": 0.0015, "step": 4605 }, { "epoch": 0.66, "learning_rate": 4.6321820658706626e-05, "loss": 0.0016, "step": 4610 }, { "epoch": 0.67, "learning_rate": 4.6311803830435134e-05, "loss": 0.0016, "step": 4615 }, { "epoch": 0.67, "learning_rate": 4.6301787002163636e-05, "loss": 0.0013, "step": 4620 }, { "epoch": 0.67, "learning_rate": 4.629177017389214e-05, "loss": 0.0011, "step": 4625 }, { "epoch": 0.67, "learning_rate": 4.628175334562064e-05, "loss": 0.0015, "step": 4630 }, { "epoch": 0.67, "learning_rate": 4.627173651734915e-05, "loss": 0.0012, "step": 4635 }, { "epoch": 0.67, "learning_rate": 4.626171968907766e-05, "loss": 0.0015, "step": 4640 }, { "epoch": 0.67, "learning_rate": 4.625170286080616e-05, "loss": 0.0014, "step": 4645 }, { "epoch": 0.67, "learning_rate": 4.624168603253466e-05, "loss": 0.0012, "step": 4650 }, { "epoch": 0.67, "learning_rate": 4.623166920426316e-05, "loss": 0.0013, "step": 4655 }, { "epoch": 0.67, "learning_rate": 4.622165237599167e-05, "loss": 0.0012, "step": 4660 }, { "epoch": 0.67, "learning_rate": 4.621163554772017e-05, "loss": 0.0016, "step": 4665 }, { "epoch": 0.67, "learning_rate": 4.6201618719448675e-05, "loss": 0.0014, "step": 4670 }, { "epoch": 0.67, "learning_rate": 4.619160189117718e-05, "loss": 0.0017, "step": 4675 }, { "epoch": 0.68, "learning_rate": 4.6181585062905685e-05, "loss": 0.0016, "step": 4680 }, { "epoch": 0.68, "learning_rate": 4.617156823463419e-05, "loss": 0.0015, "step": 4685 }, { "epoch": 0.68, "learning_rate": 4.616155140636269e-05, "loss": 0.0009, "step": 4690 }, { "epoch": 0.68, "learning_rate": 4.615153457809119e-05, "loss": 0.0018, "step": 4695 }, { "epoch": 0.68, "learning_rate": 4.61415177498197e-05, "loss": 0.0013, "step": 4700 }, { "epoch": 0.68, "learning_rate": 4.61315009215482e-05, "loss": 0.0016, "step": 4705 }, { "epoch": 0.68, "learning_rate": 4.612148409327671e-05, "loss": 0.0014, "step": 4710 }, { "epoch": 0.68, "learning_rate": 4.611146726500521e-05, "loss": 0.0014, "step": 4715 }, { "epoch": 0.68, "learning_rate": 4.610145043673372e-05, "loss": 0.0016, "step": 4720 }, { "epoch": 0.68, "learning_rate": 4.609143360846222e-05, "loss": 0.0015, "step": 4725 }, { "epoch": 0.68, "learning_rate": 4.6081416780190724e-05, "loss": 0.0013, "step": 4730 }, { "epoch": 0.68, "learning_rate": 4.6071399951919226e-05, "loss": 0.0014, "step": 4735 }, { "epoch": 0.68, "learning_rate": 4.606138312364773e-05, "loss": 0.0014, "step": 4740 }, { "epoch": 0.68, "learning_rate": 4.6051366295376236e-05, "loss": 0.0009, "step": 4745 }, { "epoch": 0.69, "learning_rate": 4.604134946710474e-05, "loss": 0.0014, "step": 4750 }, { "epoch": 0.69, "learning_rate": 4.603133263883324e-05, "loss": 0.0017, "step": 4755 }, { "epoch": 0.69, "learning_rate": 4.602131581056174e-05, "loss": 0.0007, "step": 4760 }, { "epoch": 0.69, "learning_rate": 4.601129898229025e-05, "loss": 0.0015, "step": 4765 }, { "epoch": 0.69, "learning_rate": 4.600128215401875e-05, "loss": 0.0012, "step": 4770 }, { "epoch": 0.69, "learning_rate": 4.5991265325747254e-05, "loss": 0.0019, "step": 4775 }, { "epoch": 0.69, "learning_rate": 4.598124849747576e-05, "loss": 0.0011, "step": 4780 }, { "epoch": 0.69, "learning_rate": 4.5971231669204265e-05, "loss": 0.0014, "step": 4785 }, { "epoch": 0.69, "learning_rate": 4.596121484093277e-05, "loss": 0.0014, "step": 4790 }, { "epoch": 0.69, "learning_rate": 4.5951198012661275e-05, "loss": 0.0013, "step": 4795 }, { "epoch": 0.69, "learning_rate": 4.594118118438978e-05, "loss": 0.0014, "step": 4800 }, { "epoch": 0.69, "learning_rate": 4.5931164356118285e-05, "loss": 0.0015, "step": 4805 }, { "epoch": 0.69, "learning_rate": 4.592114752784679e-05, "loss": 0.0014, "step": 4810 }, { "epoch": 0.69, "learning_rate": 4.591113069957529e-05, "loss": 0.001, "step": 4815 }, { "epoch": 0.7, "learning_rate": 4.590111387130379e-05, "loss": 0.0012, "step": 4820 }, { "epoch": 0.7, "learning_rate": 4.589109704303229e-05, "loss": 0.0011, "step": 4825 }, { "epoch": 0.7, "learning_rate": 4.58810802147608e-05, "loss": 0.0016, "step": 4830 }, { "epoch": 0.7, "learning_rate": 4.58710633864893e-05, "loss": 0.0014, "step": 4835 }, { "epoch": 0.7, "learning_rate": 4.5861046558217805e-05, "loss": 0.0016, "step": 4840 }, { "epoch": 0.7, "learning_rate": 4.585102972994631e-05, "loss": 0.0012, "step": 4845 }, { "epoch": 0.7, "learning_rate": 4.5841012901674816e-05, "loss": 0.0013, "step": 4850 }, { "epoch": 0.7, "learning_rate": 4.583099607340332e-05, "loss": 0.0015, "step": 4855 }, { "epoch": 0.7, "learning_rate": 4.5820979245131826e-05, "loss": 0.0016, "step": 4860 }, { "epoch": 0.7, "learning_rate": 4.581096241686033e-05, "loss": 0.0015, "step": 4865 }, { "epoch": 0.7, "learning_rate": 4.5800945588588836e-05, "loss": 0.0015, "step": 4870 }, { "epoch": 0.7, "learning_rate": 4.579092876031734e-05, "loss": 0.0016, "step": 4875 }, { "epoch": 0.7, "learning_rate": 4.578091193204584e-05, "loss": 0.0013, "step": 4880 }, { "epoch": 0.7, "learning_rate": 4.577089510377434e-05, "loss": 0.0013, "step": 4885 }, { "epoch": 0.71, "learning_rate": 4.5760878275502844e-05, "loss": 0.0013, "step": 4890 }, { "epoch": 0.71, "learning_rate": 4.575086144723135e-05, "loss": 0.0012, "step": 4895 }, { "epoch": 0.71, "learning_rate": 4.5740844618959854e-05, "loss": 0.0018, "step": 4900 }, { "epoch": 0.71, "learning_rate": 4.5730827790688356e-05, "loss": 0.0008, "step": 4905 }, { "epoch": 0.71, "learning_rate": 4.572081096241686e-05, "loss": 0.0012, "step": 4910 }, { "epoch": 0.71, "learning_rate": 4.5710794134145367e-05, "loss": 0.0019, "step": 4915 }, { "epoch": 0.71, "learning_rate": 4.570077730587387e-05, "loss": 0.0014, "step": 4920 }, { "epoch": 0.71, "learning_rate": 4.569076047760237e-05, "loss": 0.0009, "step": 4925 }, { "epoch": 0.71, "learning_rate": 4.568074364933088e-05, "loss": 0.0014, "step": 4930 }, { "epoch": 0.71, "learning_rate": 4.567072682105939e-05, "loss": 0.0012, "step": 4935 }, { "epoch": 0.71, "learning_rate": 4.566070999278789e-05, "loss": 0.0011, "step": 4940 }, { "epoch": 0.71, "learning_rate": 4.565069316451639e-05, "loss": 0.0013, "step": 4945 }, { "epoch": 0.71, "learning_rate": 4.564067633624489e-05, "loss": 0.0015, "step": 4950 }, { "epoch": 0.71, "learning_rate": 4.56306595079734e-05, "loss": 0.0014, "step": 4955 }, { "epoch": 0.72, "learning_rate": 4.5620642679701903e-05, "loss": 0.0014, "step": 4960 }, { "epoch": 0.72, "learning_rate": 4.5610625851430405e-05, "loss": 0.0016, "step": 4965 }, { "epoch": 0.72, "learning_rate": 4.560060902315891e-05, "loss": 0.0013, "step": 4970 }, { "epoch": 0.72, "learning_rate": 4.559059219488741e-05, "loss": 0.001, "step": 4975 }, { "epoch": 0.72, "learning_rate": 4.558057536661592e-05, "loss": 0.0014, "step": 4980 }, { "epoch": 0.72, "learning_rate": 4.557055853834442e-05, "loss": 0.0009, "step": 4985 }, { "epoch": 0.72, "learning_rate": 4.556054171007292e-05, "loss": 0.0013, "step": 4990 }, { "epoch": 0.72, "learning_rate": 4.555052488180142e-05, "loss": 0.0019, "step": 4995 }, { "epoch": 0.72, "learning_rate": 4.554050805352993e-05, "loss": 0.0012, "step": 5000 }, { "epoch": 0.72, "learning_rate": 4.553049122525844e-05, "loss": 0.0015, "step": 5005 }, { "epoch": 0.72, "learning_rate": 4.552047439698694e-05, "loss": 0.0014, "step": 5010 }, { "epoch": 0.72, "learning_rate": 4.5510457568715444e-05, "loss": 0.0014, "step": 5015 }, { "epoch": 0.72, "learning_rate": 4.550044074044395e-05, "loss": 0.0011, "step": 5020 }, { "epoch": 0.72, "learning_rate": 4.5490423912172454e-05, "loss": 0.0013, "step": 5025 }, { "epoch": 0.73, "learning_rate": 4.5480407083900956e-05, "loss": 0.0014, "step": 5030 }, { "epoch": 0.73, "learning_rate": 4.547039025562946e-05, "loss": 0.001, "step": 5035 }, { "epoch": 0.73, "learning_rate": 4.546037342735797e-05, "loss": 0.0012, "step": 5040 }, { "epoch": 0.73, "learning_rate": 4.545035659908647e-05, "loss": 0.0015, "step": 5045 }, { "epoch": 0.73, "learning_rate": 4.544033977081497e-05, "loss": 0.0009, "step": 5050 }, { "epoch": 0.73, "learning_rate": 4.543032294254347e-05, "loss": 0.0014, "step": 5055 }, { "epoch": 0.73, "learning_rate": 4.5420306114271974e-05, "loss": 0.0014, "step": 5060 }, { "epoch": 0.73, "learning_rate": 4.541028928600048e-05, "loss": 0.0022, "step": 5065 }, { "epoch": 0.73, "learning_rate": 4.5400272457728985e-05, "loss": 0.0015, "step": 5070 }, { "epoch": 0.73, "learning_rate": 4.539025562945749e-05, "loss": 0.0011, "step": 5075 }, { "epoch": 0.73, "learning_rate": 4.5380238801185995e-05, "loss": 0.001, "step": 5080 }, { "epoch": 0.73, "learning_rate": 4.5370221972914504e-05, "loss": 0.0018, "step": 5085 }, { "epoch": 0.73, "learning_rate": 4.5360205144643005e-05, "loss": 0.0012, "step": 5090 }, { "epoch": 0.73, "learning_rate": 4.535018831637151e-05, "loss": 0.0013, "step": 5095 }, { "epoch": 0.74, "learning_rate": 4.534017148810001e-05, "loss": 0.0018, "step": 5100 }, { "epoch": 0.74, "learning_rate": 4.533015465982852e-05, "loss": 0.0013, "step": 5105 }, { "epoch": 0.74, "learning_rate": 4.532013783155702e-05, "loss": 0.0015, "step": 5110 }, { "epoch": 0.74, "learning_rate": 4.531012100328552e-05, "loss": 0.0016, "step": 5115 }, { "epoch": 0.74, "learning_rate": 4.530010417501402e-05, "loss": 0.0015, "step": 5120 }, { "epoch": 0.74, "learning_rate": 4.5290087346742525e-05, "loss": 0.0012, "step": 5125 }, { "epoch": 0.74, "learning_rate": 4.5280070518471034e-05, "loss": 0.0011, "step": 5130 }, { "epoch": 0.74, "learning_rate": 4.5270053690199536e-05, "loss": 0.0015, "step": 5135 }, { "epoch": 0.74, "learning_rate": 4.526003686192804e-05, "loss": 0.0012, "step": 5140 }, { "epoch": 0.74, "learning_rate": 4.5250020033656546e-05, "loss": 0.0015, "step": 5145 }, { "epoch": 0.74, "learning_rate": 4.524000320538505e-05, "loss": 0.0013, "step": 5150 }, { "epoch": 0.74, "learning_rate": 4.5229986377113556e-05, "loss": 0.0018, "step": 5155 }, { "epoch": 0.74, "learning_rate": 4.521996954884206e-05, "loss": 0.0011, "step": 5160 }, { "epoch": 0.74, "learning_rate": 4.520995272057056e-05, "loss": 0.0011, "step": 5165 }, { "epoch": 0.75, "learning_rate": 4.519993589229907e-05, "loss": 0.001, "step": 5170 }, { "epoch": 0.75, "learning_rate": 4.518991906402757e-05, "loss": 0.0014, "step": 5175 }, { "epoch": 0.75, "learning_rate": 4.517990223575607e-05, "loss": 0.001, "step": 5180 }, { "epoch": 0.75, "learning_rate": 4.5169885407484574e-05, "loss": 0.0013, "step": 5185 }, { "epoch": 0.75, "learning_rate": 4.515986857921308e-05, "loss": 0.0011, "step": 5190 }, { "epoch": 0.75, "learning_rate": 4.5149851750941585e-05, "loss": 0.001, "step": 5195 }, { "epoch": 0.75, "learning_rate": 4.5139834922670087e-05, "loss": 0.0011, "step": 5200 }, { "epoch": 0.75, "learning_rate": 4.512981809439859e-05, "loss": 0.0012, "step": 5205 }, { "epoch": 0.75, "learning_rate": 4.511980126612709e-05, "loss": 0.0009, "step": 5210 }, { "epoch": 0.75, "learning_rate": 4.51097844378556e-05, "loss": 0.0016, "step": 5215 }, { "epoch": 0.75, "learning_rate": 4.50997676095841e-05, "loss": 0.0009, "step": 5220 }, { "epoch": 0.75, "learning_rate": 4.508975078131261e-05, "loss": 0.0016, "step": 5225 }, { "epoch": 0.75, "learning_rate": 4.507973395304111e-05, "loss": 0.0008, "step": 5230 }, { "epoch": 0.76, "learning_rate": 4.506971712476962e-05, "loss": 0.0012, "step": 5235 }, { "epoch": 0.76, "learning_rate": 4.505970029649812e-05, "loss": 0.001, "step": 5240 }, { "epoch": 0.76, "learning_rate": 4.5049683468226623e-05, "loss": 0.0012, "step": 5245 }, { "epoch": 0.76, "learning_rate": 4.5039666639955125e-05, "loss": 0.0016, "step": 5250 }, { "epoch": 0.76, "learning_rate": 4.5029649811683634e-05, "loss": 0.0012, "step": 5255 }, { "epoch": 0.76, "learning_rate": 4.5019632983412136e-05, "loss": 0.0012, "step": 5260 }, { "epoch": 0.76, "learning_rate": 4.500961615514064e-05, "loss": 0.0011, "step": 5265 }, { "epoch": 0.76, "learning_rate": 4.499959932686914e-05, "loss": 0.0016, "step": 5270 }, { "epoch": 0.76, "learning_rate": 4.498958249859765e-05, "loss": 0.0012, "step": 5275 }, { "epoch": 0.76, "learning_rate": 4.497956567032615e-05, "loss": 0.0012, "step": 5280 }, { "epoch": 0.76, "learning_rate": 4.496954884205465e-05, "loss": 0.0016, "step": 5285 }, { "epoch": 0.76, "learning_rate": 4.4959532013783154e-05, "loss": 0.0009, "step": 5290 }, { "epoch": 0.76, "learning_rate": 4.494951518551166e-05, "loss": 0.0013, "step": 5295 }, { "epoch": 0.76, "learning_rate": 4.4939498357240164e-05, "loss": 0.0008, "step": 5300 }, { "epoch": 0.77, "learning_rate": 4.492948152896867e-05, "loss": 0.0014, "step": 5305 }, { "epoch": 0.77, "learning_rate": 4.4919464700697174e-05, "loss": 0.002, "step": 5310 }, { "epoch": 0.77, "learning_rate": 4.4909447872425676e-05, "loss": 0.0015, "step": 5315 }, { "epoch": 0.77, "learning_rate": 4.4899431044154185e-05, "loss": 0.0018, "step": 5320 }, { "epoch": 0.77, "learning_rate": 4.488941421588269e-05, "loss": 0.0017, "step": 5325 }, { "epoch": 0.77, "learning_rate": 4.487939738761119e-05, "loss": 0.0011, "step": 5330 }, { "epoch": 0.77, "learning_rate": 4.486938055933969e-05, "loss": 0.0018, "step": 5335 }, { "epoch": 0.77, "learning_rate": 4.48593637310682e-05, "loss": 0.001, "step": 5340 }, { "epoch": 0.77, "learning_rate": 4.48493469027967e-05, "loss": 0.0018, "step": 5345 }, { "epoch": 0.77, "learning_rate": 4.48393300745252e-05, "loss": 0.0006, "step": 5350 }, { "epoch": 0.77, "learning_rate": 4.4829313246253705e-05, "loss": 0.0011, "step": 5355 }, { "epoch": 0.77, "learning_rate": 4.481929641798221e-05, "loss": 0.0014, "step": 5360 }, { "epoch": 0.77, "learning_rate": 4.4809279589710715e-05, "loss": 0.001, "step": 5365 }, { "epoch": 0.77, "learning_rate": 4.479926276143922e-05, "loss": 0.0009, "step": 5370 }, { "epoch": 0.78, "learning_rate": 4.4789245933167725e-05, "loss": 0.0016, "step": 5375 }, { "epoch": 0.78, "learning_rate": 4.477922910489623e-05, "loss": 0.0012, "step": 5380 }, { "epoch": 0.78, "learning_rate": 4.4769212276624736e-05, "loss": 0.0009, "step": 5385 }, { "epoch": 0.78, "learning_rate": 4.475919544835324e-05, "loss": 0.0012, "step": 5390 }, { "epoch": 0.78, "learning_rate": 4.474917862008174e-05, "loss": 0.0014, "step": 5395 }, { "epoch": 0.78, "learning_rate": 4.473916179181024e-05, "loss": 0.0014, "step": 5400 }, { "epoch": 0.78, "learning_rate": 4.472914496353875e-05, "loss": 0.0014, "step": 5405 }, { "epoch": 0.78, "learning_rate": 4.471912813526725e-05, "loss": 0.0018, "step": 5410 }, { "epoch": 0.78, "learning_rate": 4.4709111306995754e-05, "loss": 0.0015, "step": 5415 }, { "epoch": 0.78, "learning_rate": 4.4699094478724256e-05, "loss": 0.0013, "step": 5420 }, { "epoch": 0.78, "learning_rate": 4.4689077650452764e-05, "loss": 0.0013, "step": 5425 }, { "epoch": 0.78, "learning_rate": 4.4679060822181266e-05, "loss": 0.0014, "step": 5430 }, { "epoch": 0.78, "learning_rate": 4.466904399390977e-05, "loss": 0.0011, "step": 5435 }, { "epoch": 0.78, "learning_rate": 4.465902716563827e-05, "loss": 0.0014, "step": 5440 }, { "epoch": 0.79, "learning_rate": 4.464901033736678e-05, "loss": 0.0015, "step": 5445 }, { "epoch": 0.79, "learning_rate": 4.463899350909529e-05, "loss": 0.001, "step": 5450 }, { "epoch": 0.79, "learning_rate": 4.462897668082379e-05, "loss": 0.0011, "step": 5455 }, { "epoch": 0.79, "learning_rate": 4.461895985255229e-05, "loss": 0.0009, "step": 5460 }, { "epoch": 0.79, "learning_rate": 4.460894302428079e-05, "loss": 0.0014, "step": 5465 }, { "epoch": 0.79, "learning_rate": 4.45989261960093e-05, "loss": 0.0017, "step": 5470 }, { "epoch": 0.79, "learning_rate": 4.45889093677378e-05, "loss": 0.0013, "step": 5475 }, { "epoch": 0.79, "learning_rate": 4.4578892539466305e-05, "loss": 0.001, "step": 5480 }, { "epoch": 0.79, "learning_rate": 4.4568875711194807e-05, "loss": 0.0011, "step": 5485 }, { "epoch": 0.79, "learning_rate": 4.4558858882923315e-05, "loss": 0.0015, "step": 5490 }, { "epoch": 0.79, "learning_rate": 4.454884205465182e-05, "loss": 0.001, "step": 5495 }, { "epoch": 0.79, "learning_rate": 4.453882522638032e-05, "loss": 0.0011, "step": 5500 }, { "epoch": 0.79, "learning_rate": 4.452880839810882e-05, "loss": 0.0014, "step": 5505 }, { "epoch": 0.79, "learning_rate": 4.451879156983733e-05, "loss": 0.0014, "step": 5510 }, { "epoch": 0.8, "learning_rate": 4.450877474156583e-05, "loss": 0.0013, "step": 5515 }, { "epoch": 0.8, "learning_rate": 4.449875791329434e-05, "loss": 0.0007, "step": 5520 }, { "epoch": 0.8, "learning_rate": 4.448874108502284e-05, "loss": 0.0014, "step": 5525 }, { "epoch": 0.8, "learning_rate": 4.4478724256751343e-05, "loss": 0.0014, "step": 5530 }, { "epoch": 0.8, "learning_rate": 4.446870742847985e-05, "loss": 0.0017, "step": 5535 }, { "epoch": 0.8, "learning_rate": 4.4458690600208354e-05, "loss": 0.0011, "step": 5540 }, { "epoch": 0.8, "learning_rate": 4.4448673771936856e-05, "loss": 0.0007, "step": 5545 }, { "epoch": 0.8, "learning_rate": 4.443865694366536e-05, "loss": 0.0016, "step": 5550 }, { "epoch": 0.8, "learning_rate": 4.4428640115393866e-05, "loss": 0.0011, "step": 5555 }, { "epoch": 0.8, "learning_rate": 4.441862328712237e-05, "loss": 0.0012, "step": 5560 }, { "epoch": 0.8, "learning_rate": 4.440860645885087e-05, "loss": 0.0013, "step": 5565 }, { "epoch": 0.8, "learning_rate": 4.439858963057937e-05, "loss": 0.0012, "step": 5570 }, { "epoch": 0.8, "learning_rate": 4.438857280230788e-05, "loss": 0.0017, "step": 5575 }, { "epoch": 0.8, "learning_rate": 4.437855597403638e-05, "loss": 0.0016, "step": 5580 }, { "epoch": 0.81, "learning_rate": 4.4368539145764884e-05, "loss": 0.0014, "step": 5585 }, { "epoch": 0.81, "learning_rate": 4.435852231749339e-05, "loss": 0.0012, "step": 5590 }, { "epoch": 0.81, "learning_rate": 4.4348505489221894e-05, "loss": 0.0012, "step": 5595 }, { "epoch": 0.81, "learning_rate": 4.43384886609504e-05, "loss": 0.0007, "step": 5600 }, { "epoch": 0.81, "learning_rate": 4.4328471832678905e-05, "loss": 0.0008, "step": 5605 }, { "epoch": 0.81, "learning_rate": 4.431845500440741e-05, "loss": 0.001, "step": 5610 }, { "epoch": 0.81, "learning_rate": 4.430843817613591e-05, "loss": 0.0009, "step": 5615 }, { "epoch": 0.81, "learning_rate": 4.429842134786442e-05, "loss": 0.0009, "step": 5620 }, { "epoch": 0.81, "learning_rate": 4.428840451959292e-05, "loss": 0.0012, "step": 5625 }, { "epoch": 0.81, "learning_rate": 4.427838769132142e-05, "loss": 0.001, "step": 5630 }, { "epoch": 0.81, "learning_rate": 4.426837086304992e-05, "loss": 0.0007, "step": 5635 }, { "epoch": 0.81, "learning_rate": 4.425835403477843e-05, "loss": 0.0011, "step": 5640 }, { "epoch": 0.81, "learning_rate": 4.424833720650693e-05, "loss": 0.0007, "step": 5645 }, { "epoch": 0.81, "learning_rate": 4.4238320378235435e-05, "loss": 0.0018, "step": 5650 }, { "epoch": 0.82, "learning_rate": 4.422830354996394e-05, "loss": 0.0008, "step": 5655 }, { "epoch": 0.82, "learning_rate": 4.4218286721692445e-05, "loss": 0.0011, "step": 5660 }, { "epoch": 0.82, "learning_rate": 4.420826989342095e-05, "loss": 0.0013, "step": 5665 }, { "epoch": 0.82, "learning_rate": 4.4198253065149456e-05, "loss": 0.0011, "step": 5670 }, { "epoch": 0.82, "learning_rate": 4.418823623687796e-05, "loss": 0.0018, "step": 5675 }, { "epoch": 0.82, "learning_rate": 4.417821940860646e-05, "loss": 0.0012, "step": 5680 }, { "epoch": 0.82, "learning_rate": 4.416820258033497e-05, "loss": 0.0014, "step": 5685 }, { "epoch": 0.82, "learning_rate": 4.415818575206347e-05, "loss": 0.001, "step": 5690 }, { "epoch": 0.82, "learning_rate": 4.414816892379197e-05, "loss": 0.001, "step": 5695 }, { "epoch": 0.82, "learning_rate": 4.4138152095520474e-05, "loss": 0.0012, "step": 5700 }, { "epoch": 0.82, "learning_rate": 4.412813526724898e-05, "loss": 0.0014, "step": 5705 }, { "epoch": 0.82, "learning_rate": 4.4118118438977484e-05, "loss": 0.001, "step": 5710 }, { "epoch": 0.82, "learning_rate": 4.4108101610705986e-05, "loss": 0.0013, "step": 5715 }, { "epoch": 0.83, "learning_rate": 4.409808478243449e-05, "loss": 0.0009, "step": 5720 }, { "epoch": 0.83, "learning_rate": 4.4088067954162996e-05, "loss": 0.0007, "step": 5725 }, { "epoch": 0.83, "learning_rate": 4.40780511258915e-05, "loss": 0.0008, "step": 5730 }, { "epoch": 0.83, "learning_rate": 4.406803429762e-05, "loss": 0.0009, "step": 5735 }, { "epoch": 0.83, "learning_rate": 4.405801746934851e-05, "loss": 0.0013, "step": 5740 }, { "epoch": 0.83, "learning_rate": 4.404800064107702e-05, "loss": 0.0017, "step": 5745 }, { "epoch": 0.83, "learning_rate": 4.403798381280552e-05, "loss": 0.0012, "step": 5750 }, { "epoch": 0.83, "learning_rate": 4.402796698453402e-05, "loss": 0.0011, "step": 5755 }, { "epoch": 0.83, "learning_rate": 4.401795015626252e-05, "loss": 0.001, "step": 5760 }, { "epoch": 0.83, "learning_rate": 4.4007933327991025e-05, "loss": 0.001, "step": 5765 }, { "epoch": 0.83, "learning_rate": 4.399791649971953e-05, "loss": 0.001, "step": 5770 }, { "epoch": 0.83, "learning_rate": 4.3987899671448035e-05, "loss": 0.0016, "step": 5775 }, { "epoch": 0.83, "learning_rate": 4.397788284317654e-05, "loss": 0.0013, "step": 5780 }, { "epoch": 0.83, "learning_rate": 4.396786601490504e-05, "loss": 0.0006, "step": 5785 }, { "epoch": 0.84, "learning_rate": 4.395784918663355e-05, "loss": 0.0014, "step": 5790 }, { "epoch": 0.84, "learning_rate": 4.394783235836205e-05, "loss": 0.0016, "step": 5795 }, { "epoch": 0.84, "learning_rate": 4.393781553009055e-05, "loss": 0.0011, "step": 5800 }, { "epoch": 0.84, "learning_rate": 4.392779870181905e-05, "loss": 0.0007, "step": 5805 }, { "epoch": 0.84, "learning_rate": 4.391778187354756e-05, "loss": 0.0008, "step": 5810 }, { "epoch": 0.84, "learning_rate": 4.390776504527607e-05, "loss": 0.0008, "step": 5815 }, { "epoch": 0.84, "learning_rate": 4.389774821700457e-05, "loss": 0.0008, "step": 5820 }, { "epoch": 0.84, "learning_rate": 4.3887731388733074e-05, "loss": 0.0011, "step": 5825 }, { "epoch": 0.84, "learning_rate": 4.387771456046158e-05, "loss": 0.001, "step": 5830 }, { "epoch": 0.84, "learning_rate": 4.3867697732190084e-05, "loss": 0.0014, "step": 5835 }, { "epoch": 0.84, "learning_rate": 4.3857680903918586e-05, "loss": 0.0016, "step": 5840 }, { "epoch": 0.84, "learning_rate": 4.384766407564709e-05, "loss": 0.0013, "step": 5845 }, { "epoch": 0.84, "learning_rate": 4.383764724737559e-05, "loss": 0.0015, "step": 5850 }, { "epoch": 0.84, "learning_rate": 4.38276304191041e-05, "loss": 0.0011, "step": 5855 }, { "epoch": 0.85, "learning_rate": 4.38176135908326e-05, "loss": 0.0013, "step": 5860 }, { "epoch": 0.85, "learning_rate": 4.38075967625611e-05, "loss": 0.0016, "step": 5865 }, { "epoch": 0.85, "learning_rate": 4.3797579934289604e-05, "loss": 0.001, "step": 5870 }, { "epoch": 0.85, "learning_rate": 4.378756310601811e-05, "loss": 0.0013, "step": 5875 }, { "epoch": 0.85, "learning_rate": 4.3777546277746614e-05, "loss": 0.0009, "step": 5880 }, { "epoch": 0.85, "learning_rate": 4.376752944947512e-05, "loss": 0.0012, "step": 5885 }, { "epoch": 0.85, "learning_rate": 4.3757512621203625e-05, "loss": 0.0011, "step": 5890 }, { "epoch": 0.85, "learning_rate": 4.3747495792932133e-05, "loss": 0.0014, "step": 5895 }, { "epoch": 0.85, "learning_rate": 4.3737478964660635e-05, "loss": 0.0012, "step": 5900 }, { "epoch": 0.85, "learning_rate": 4.372746213638914e-05, "loss": 0.0011, "step": 5905 }, { "epoch": 0.85, "learning_rate": 4.371744530811764e-05, "loss": 0.0013, "step": 5910 }, { "epoch": 0.85, "learning_rate": 4.370742847984614e-05, "loss": 0.001, "step": 5915 }, { "epoch": 0.85, "learning_rate": 4.369741165157465e-05, "loss": 0.0013, "step": 5920 }, { "epoch": 0.85, "learning_rate": 4.368739482330315e-05, "loss": 0.001, "step": 5925 }, { "epoch": 0.86, "learning_rate": 4.367737799503165e-05, "loss": 0.0007, "step": 5930 }, { "epoch": 0.86, "learning_rate": 4.3667361166760155e-05, "loss": 0.0012, "step": 5935 }, { "epoch": 0.86, "learning_rate": 4.3657344338488664e-05, "loss": 0.0014, "step": 5940 }, { "epoch": 0.86, "learning_rate": 4.3647327510217165e-05, "loss": 0.0008, "step": 5945 }, { "epoch": 0.86, "learning_rate": 4.363731068194567e-05, "loss": 0.0011, "step": 5950 }, { "epoch": 0.86, "learning_rate": 4.3627293853674176e-05, "loss": 0.0014, "step": 5955 }, { "epoch": 0.86, "learning_rate": 4.361727702540268e-05, "loss": 0.0012, "step": 5960 }, { "epoch": 0.86, "learning_rate": 4.3607260197131186e-05, "loss": 0.0013, "step": 5965 }, { "epoch": 0.86, "learning_rate": 4.359724336885969e-05, "loss": 0.0009, "step": 5970 }, { "epoch": 0.86, "learning_rate": 4.358722654058819e-05, "loss": 0.0018, "step": 5975 }, { "epoch": 0.86, "learning_rate": 4.35772097123167e-05, "loss": 0.0013, "step": 5980 }, { "epoch": 0.86, "learning_rate": 4.35671928840452e-05, "loss": 0.0009, "step": 5985 }, { "epoch": 0.86, "learning_rate": 4.35571760557737e-05, "loss": 0.0012, "step": 5990 }, { "epoch": 0.86, "learning_rate": 4.3547159227502204e-05, "loss": 0.0015, "step": 5995 }, { "epoch": 0.87, "learning_rate": 4.3537142399230706e-05, "loss": 0.0014, "step": 6000 }, { "epoch": 0.87, "learning_rate": 4.3527125570959215e-05, "loss": 0.0013, "step": 6005 }, { "epoch": 0.87, "learning_rate": 4.3517108742687716e-05, "loss": 0.001, "step": 6010 }, { "epoch": 0.87, "learning_rate": 4.350709191441622e-05, "loss": 0.0008, "step": 6015 }, { "epoch": 0.87, "learning_rate": 4.349707508614472e-05, "loss": 0.0014, "step": 6020 }, { "epoch": 0.87, "learning_rate": 4.348705825787323e-05, "loss": 0.0011, "step": 6025 }, { "epoch": 0.87, "learning_rate": 4.347704142960173e-05, "loss": 0.0015, "step": 6030 }, { "epoch": 0.87, "learning_rate": 4.346702460133024e-05, "loss": 0.0008, "step": 6035 }, { "epoch": 0.87, "learning_rate": 4.345700777305874e-05, "loss": 0.0011, "step": 6040 }, { "epoch": 0.87, "learning_rate": 4.344699094478725e-05, "loss": 0.0015, "step": 6045 }, { "epoch": 0.87, "learning_rate": 4.343697411651575e-05, "loss": 0.0012, "step": 6050 }, { "epoch": 0.87, "learning_rate": 4.342695728824425e-05, "loss": 0.0011, "step": 6055 }, { "epoch": 0.87, "learning_rate": 4.3416940459972755e-05, "loss": 0.0018, "step": 6060 }, { "epoch": 0.87, "learning_rate": 4.3406923631701264e-05, "loss": 0.0008, "step": 6065 }, { "epoch": 0.88, "learning_rate": 4.3396906803429766e-05, "loss": 0.001, "step": 6070 }, { "epoch": 0.88, "learning_rate": 4.338688997515827e-05, "loss": 0.0013, "step": 6075 }, { "epoch": 0.88, "learning_rate": 4.337687314688677e-05, "loss": 0.0012, "step": 6080 }, { "epoch": 0.88, "learning_rate": 4.336685631861527e-05, "loss": 0.0015, "step": 6085 }, { "epoch": 0.88, "learning_rate": 4.335683949034378e-05, "loss": 0.001, "step": 6090 }, { "epoch": 0.88, "learning_rate": 4.334682266207228e-05, "loss": 0.0013, "step": 6095 }, { "epoch": 0.88, "learning_rate": 4.3336805833800783e-05, "loss": 0.0016, "step": 6100 }, { "epoch": 0.88, "learning_rate": 4.332678900552929e-05, "loss": 0.0011, "step": 6105 }, { "epoch": 0.88, "learning_rate": 4.3316772177257794e-05, "loss": 0.0016, "step": 6110 }, { "epoch": 0.88, "learning_rate": 4.33067553489863e-05, "loss": 0.0008, "step": 6115 }, { "epoch": 0.88, "learning_rate": 4.3296738520714804e-05, "loss": 0.001, "step": 6120 }, { "epoch": 0.88, "learning_rate": 4.3286721692443306e-05, "loss": 0.0013, "step": 6125 }, { "epoch": 0.88, "learning_rate": 4.3276704864171815e-05, "loss": 0.0012, "step": 6130 }, { "epoch": 0.88, "learning_rate": 4.3266688035900317e-05, "loss": 0.0012, "step": 6135 }, { "epoch": 0.89, "learning_rate": 4.325667120762882e-05, "loss": 0.0017, "step": 6140 }, { "epoch": 0.89, "learning_rate": 4.324665437935732e-05, "loss": 0.0013, "step": 6145 }, { "epoch": 0.89, "learning_rate": 4.323663755108583e-05, "loss": 0.0012, "step": 6150 }, { "epoch": 0.89, "learning_rate": 4.322662072281433e-05, "loss": 0.0012, "step": 6155 }, { "epoch": 0.89, "learning_rate": 4.321660389454283e-05, "loss": 0.0013, "step": 6160 }, { "epoch": 0.89, "learning_rate": 4.3206587066271334e-05, "loss": 0.0011, "step": 6165 }, { "epoch": 0.89, "learning_rate": 4.3196570237999836e-05, "loss": 0.0009, "step": 6170 }, { "epoch": 0.89, "learning_rate": 4.3186553409728345e-05, "loss": 0.0018, "step": 6175 }, { "epoch": 0.89, "learning_rate": 4.3176536581456853e-05, "loss": 0.0015, "step": 6180 }, { "epoch": 0.89, "learning_rate": 4.3166519753185355e-05, "loss": 0.0011, "step": 6185 }, { "epoch": 0.89, "learning_rate": 4.315650292491386e-05, "loss": 0.0013, "step": 6190 }, { "epoch": 0.89, "learning_rate": 4.3146486096642366e-05, "loss": 0.0013, "step": 6195 }, { "epoch": 0.89, "learning_rate": 4.313646926837087e-05, "loss": 0.0011, "step": 6200 }, { "epoch": 0.89, "learning_rate": 4.312645244009937e-05, "loss": 0.0012, "step": 6205 }, { "epoch": 0.9, "learning_rate": 4.311643561182787e-05, "loss": 0.0012, "step": 6210 }, { "epoch": 0.9, "learning_rate": 4.310641878355638e-05, "loss": 0.0014, "step": 6215 }, { "epoch": 0.9, "learning_rate": 4.309640195528488e-05, "loss": 0.0011, "step": 6220 }, { "epoch": 0.9, "learning_rate": 4.3086385127013384e-05, "loss": 0.0012, "step": 6225 }, { "epoch": 0.9, "learning_rate": 4.3076368298741885e-05, "loss": 0.0014, "step": 6230 }, { "epoch": 0.9, "learning_rate": 4.306635147047039e-05, "loss": 0.0013, "step": 6235 }, { "epoch": 0.9, "learning_rate": 4.3056334642198896e-05, "loss": 0.0015, "step": 6240 }, { "epoch": 0.9, "learning_rate": 4.30463178139274e-05, "loss": 0.0016, "step": 6245 }, { "epoch": 0.9, "learning_rate": 4.3036300985655906e-05, "loss": 0.0012, "step": 6250 }, { "epoch": 0.9, "learning_rate": 4.302628415738441e-05, "loss": 0.0015, "step": 6255 }, { "epoch": 0.9, "learning_rate": 4.301626732911292e-05, "loss": 0.0012, "step": 6260 }, { "epoch": 0.9, "learning_rate": 4.300625050084142e-05, "loss": 0.0018, "step": 6265 }, { "epoch": 0.9, "learning_rate": 4.299623367256992e-05, "loss": 0.0009, "step": 6270 }, { "epoch": 0.91, "learning_rate": 4.298621684429842e-05, "loss": 0.0012, "step": 6275 }, { "epoch": 0.91, "learning_rate": 4.297620001602693e-05, "loss": 0.0012, "step": 6280 }, { "epoch": 0.91, "learning_rate": 4.296618318775543e-05, "loss": 0.0013, "step": 6285 }, { "epoch": 0.91, "learning_rate": 4.2956166359483935e-05, "loss": 0.0011, "step": 6290 }, { "epoch": 0.91, "learning_rate": 4.2946149531212436e-05, "loss": 0.0012, "step": 6295 }, { "epoch": 0.91, "learning_rate": 4.2936132702940945e-05, "loss": 0.0014, "step": 6300 }, { "epoch": 0.91, "learning_rate": 4.292611587466945e-05, "loss": 0.0016, "step": 6305 }, { "epoch": 0.91, "learning_rate": 4.291609904639795e-05, "loss": 0.0011, "step": 6310 }, { "epoch": 0.91, "learning_rate": 4.290608221812645e-05, "loss": 0.0006, "step": 6315 }, { "epoch": 0.91, "learning_rate": 4.289606538985496e-05, "loss": 0.0011, "step": 6320 }, { "epoch": 0.91, "learning_rate": 4.288604856158346e-05, "loss": 0.0009, "step": 6325 }, { "epoch": 0.91, "learning_rate": 4.287603173331197e-05, "loss": 0.0011, "step": 6330 }, { "epoch": 0.91, "learning_rate": 4.286601490504047e-05, "loss": 0.001, "step": 6335 }, { "epoch": 0.91, "learning_rate": 4.285599807676897e-05, "loss": 0.0008, "step": 6340 }, { "epoch": 0.92, "learning_rate": 4.284598124849748e-05, "loss": 0.0015, "step": 6345 }, { "epoch": 0.92, "learning_rate": 4.2835964420225984e-05, "loss": 0.0014, "step": 6350 }, { "epoch": 0.92, "learning_rate": 4.2825947591954486e-05, "loss": 0.0011, "step": 6355 }, { "epoch": 0.92, "learning_rate": 4.281593076368299e-05, "loss": 0.0013, "step": 6360 }, { "epoch": 0.92, "learning_rate": 4.2805913935411496e-05, "loss": 0.0012, "step": 6365 }, { "epoch": 0.92, "learning_rate": 4.279589710714e-05, "loss": 0.0013, "step": 6370 }, { "epoch": 0.92, "learning_rate": 4.27858802788685e-05, "loss": 0.0009, "step": 6375 }, { "epoch": 0.92, "learning_rate": 4.2775863450597e-05, "loss": 0.0014, "step": 6380 }, { "epoch": 0.92, "learning_rate": 4.276584662232551e-05, "loss": 0.0009, "step": 6385 }, { "epoch": 0.92, "learning_rate": 4.275582979405401e-05, "loss": 0.0013, "step": 6390 }, { "epoch": 0.92, "learning_rate": 4.2745812965782514e-05, "loss": 0.0012, "step": 6395 }, { "epoch": 0.92, "learning_rate": 4.273579613751102e-05, "loss": 0.0009, "step": 6400 }, { "epoch": 0.92, "learning_rate": 4.2725779309239524e-05, "loss": 0.0011, "step": 6405 }, { "epoch": 0.92, "learning_rate": 4.271576248096803e-05, "loss": 0.001, "step": 6410 }, { "epoch": 0.93, "learning_rate": 4.2705745652696535e-05, "loss": 0.0014, "step": 6415 }, { "epoch": 0.93, "learning_rate": 4.2695728824425037e-05, "loss": 0.0006, "step": 6420 }, { "epoch": 0.93, "learning_rate": 4.268571199615354e-05, "loss": 0.0007, "step": 6425 }, { "epoch": 0.93, "learning_rate": 4.267569516788205e-05, "loss": 0.0011, "step": 6430 }, { "epoch": 0.93, "learning_rate": 4.266567833961055e-05, "loss": 0.001, "step": 6435 }, { "epoch": 0.93, "learning_rate": 4.265566151133905e-05, "loss": 0.0009, "step": 6440 }, { "epoch": 0.93, "learning_rate": 4.264564468306755e-05, "loss": 0.0011, "step": 6445 }, { "epoch": 0.93, "learning_rate": 4.263562785479606e-05, "loss": 0.0006, "step": 6450 }, { "epoch": 0.93, "learning_rate": 4.262561102652456e-05, "loss": 0.001, "step": 6455 }, { "epoch": 0.93, "learning_rate": 4.2615594198253065e-05, "loss": 0.001, "step": 6460 }, { "epoch": 0.93, "learning_rate": 4.260557736998157e-05, "loss": 0.0012, "step": 6465 }, { "epoch": 0.93, "learning_rate": 4.2595560541710075e-05, "loss": 0.0009, "step": 6470 }, { "epoch": 0.93, "learning_rate": 4.258554371343858e-05, "loss": 0.0016, "step": 6475 }, { "epoch": 0.93, "learning_rate": 4.2575526885167086e-05, "loss": 0.0011, "step": 6480 }, { "epoch": 0.94, "learning_rate": 4.256551005689559e-05, "loss": 0.0013, "step": 6485 }, { "epoch": 0.94, "learning_rate": 4.255549322862409e-05, "loss": 0.0008, "step": 6490 }, { "epoch": 0.94, "learning_rate": 4.25454764003526e-05, "loss": 0.0013, "step": 6495 }, { "epoch": 0.94, "learning_rate": 4.25354595720811e-05, "loss": 0.0014, "step": 6500 }, { "epoch": 0.94, "learning_rate": 4.25254427438096e-05, "loss": 0.0008, "step": 6505 }, { "epoch": 0.94, "learning_rate": 4.2515425915538104e-05, "loss": 0.0008, "step": 6510 }, { "epoch": 0.94, "learning_rate": 4.250540908726661e-05, "loss": 0.0008, "step": 6515 }, { "epoch": 0.94, "learning_rate": 4.2495392258995114e-05, "loss": 0.0012, "step": 6520 }, { "epoch": 0.94, "learning_rate": 4.2485375430723616e-05, "loss": 0.0012, "step": 6525 }, { "epoch": 0.94, "learning_rate": 4.247535860245212e-05, "loss": 0.0008, "step": 6530 }, { "epoch": 0.94, "learning_rate": 4.2465341774180626e-05, "loss": 0.001, "step": 6535 }, { "epoch": 0.94, "learning_rate": 4.245532494590913e-05, "loss": 0.001, "step": 6540 }, { "epoch": 0.94, "learning_rate": 4.244530811763763e-05, "loss": 0.0013, "step": 6545 }, { "epoch": 0.94, "learning_rate": 4.243529128936614e-05, "loss": 0.0008, "step": 6550 }, { "epoch": 0.95, "learning_rate": 4.242527446109464e-05, "loss": 0.0008, "step": 6555 }, { "epoch": 0.95, "learning_rate": 4.241525763282315e-05, "loss": 0.0012, "step": 6560 }, { "epoch": 0.95, "learning_rate": 4.240524080455165e-05, "loss": 0.0013, "step": 6565 }, { "epoch": 0.95, "learning_rate": 4.239522397628015e-05, "loss": 0.0013, "step": 6570 }, { "epoch": 0.95, "learning_rate": 4.2385207148008655e-05, "loss": 0.0013, "step": 6575 }, { "epoch": 0.95, "learning_rate": 4.237519031973716e-05, "loss": 0.0012, "step": 6580 }, { "epoch": 0.95, "learning_rate": 4.2365173491465665e-05, "loss": 0.001, "step": 6585 }, { "epoch": 0.95, "learning_rate": 4.235515666319417e-05, "loss": 0.0012, "step": 6590 }, { "epoch": 0.95, "learning_rate": 4.234513983492267e-05, "loss": 0.0012, "step": 6595 }, { "epoch": 0.95, "learning_rate": 4.233512300665118e-05, "loss": 0.0008, "step": 6600 }, { "epoch": 0.95, "learning_rate": 4.232510617837968e-05, "loss": 0.0011, "step": 6605 }, { "epoch": 0.95, "learning_rate": 4.231508935010818e-05, "loss": 0.0014, "step": 6610 }, { "epoch": 0.95, "learning_rate": 4.230507252183668e-05, "loss": 0.0013, "step": 6615 }, { "epoch": 0.95, "learning_rate": 4.229505569356519e-05, "loss": 0.0011, "step": 6620 }, { "epoch": 0.96, "learning_rate": 4.22850388652937e-05, "loss": 0.0008, "step": 6625 }, { "epoch": 0.96, "learning_rate": 4.22750220370222e-05, "loss": 0.0016, "step": 6630 }, { "epoch": 0.96, "learning_rate": 4.2265005208750704e-05, "loss": 0.0006, "step": 6635 }, { "epoch": 0.96, "learning_rate": 4.2254988380479206e-05, "loss": 0.0008, "step": 6640 }, { "epoch": 0.96, "learning_rate": 4.2244971552207714e-05, "loss": 0.0013, "step": 6645 }, { "epoch": 0.96, "learning_rate": 4.2234954723936216e-05, "loss": 0.0009, "step": 6650 }, { "epoch": 0.96, "learning_rate": 4.222493789566472e-05, "loss": 0.0011, "step": 6655 }, { "epoch": 0.96, "learning_rate": 4.221492106739322e-05, "loss": 0.001, "step": 6660 }, { "epoch": 0.96, "learning_rate": 4.220490423912173e-05, "loss": 0.0014, "step": 6665 }, { "epoch": 0.96, "learning_rate": 4.219488741085023e-05, "loss": 0.0012, "step": 6670 }, { "epoch": 0.96, "learning_rate": 4.218487058257873e-05, "loss": 0.001, "step": 6675 }, { "epoch": 0.96, "learning_rate": 4.2174853754307234e-05, "loss": 0.0014, "step": 6680 }, { "epoch": 0.96, "learning_rate": 4.216483692603574e-05, "loss": 0.0016, "step": 6685 }, { "epoch": 0.96, "learning_rate": 4.2154820097764244e-05, "loss": 0.0013, "step": 6690 }, { "epoch": 0.97, "learning_rate": 4.214480326949275e-05, "loss": 0.0009, "step": 6695 }, { "epoch": 0.97, "learning_rate": 4.2134786441221255e-05, "loss": 0.0011, "step": 6700 }, { "epoch": 0.97, "learning_rate": 4.2124769612949757e-05, "loss": 0.001, "step": 6705 }, { "epoch": 0.97, "learning_rate": 4.2114752784678265e-05, "loss": 0.0011, "step": 6710 }, { "epoch": 0.97, "learning_rate": 4.210473595640677e-05, "loss": 0.001, "step": 6715 }, { "epoch": 0.97, "learning_rate": 4.209471912813527e-05, "loss": 0.0009, "step": 6720 }, { "epoch": 0.97, "learning_rate": 4.208470229986377e-05, "loss": 0.0013, "step": 6725 }, { "epoch": 0.97, "learning_rate": 4.207468547159228e-05, "loss": 0.001, "step": 6730 }, { "epoch": 0.97, "learning_rate": 4.206466864332078e-05, "loss": 0.0008, "step": 6735 }, { "epoch": 0.97, "learning_rate": 4.205465181504928e-05, "loss": 0.0013, "step": 6740 }, { "epoch": 0.97, "learning_rate": 4.2044634986777785e-05, "loss": 0.001, "step": 6745 }, { "epoch": 0.97, "learning_rate": 4.2034618158506293e-05, "loss": 0.0008, "step": 6750 }, { "epoch": 0.97, "learning_rate": 4.2024601330234795e-05, "loss": 0.0015, "step": 6755 }, { "epoch": 0.98, "learning_rate": 4.20145845019633e-05, "loss": 0.001, "step": 6760 }, { "epoch": 0.98, "learning_rate": 4.2004567673691806e-05, "loss": 0.0013, "step": 6765 }, { "epoch": 0.98, "learning_rate": 4.199455084542031e-05, "loss": 0.0015, "step": 6770 }, { "epoch": 0.98, "learning_rate": 4.1984534017148816e-05, "loss": 0.0016, "step": 6775 }, { "epoch": 0.98, "learning_rate": 4.197451718887732e-05, "loss": 0.001, "step": 6780 }, { "epoch": 0.98, "learning_rate": 4.196450036060582e-05, "loss": 0.0013, "step": 6785 }, { "epoch": 0.98, "learning_rate": 4.195448353233432e-05, "loss": 0.0007, "step": 6790 }, { "epoch": 0.98, "learning_rate": 4.194446670406283e-05, "loss": 0.0008, "step": 6795 }, { "epoch": 0.98, "learning_rate": 4.193444987579133e-05, "loss": 0.0015, "step": 6800 }, { "epoch": 0.98, "learning_rate": 4.1924433047519834e-05, "loss": 0.0011, "step": 6805 }, { "epoch": 0.98, "learning_rate": 4.1914416219248336e-05, "loss": 0.0009, "step": 6810 }, { "epoch": 0.98, "learning_rate": 4.1904399390976844e-05, "loss": 0.0012, "step": 6815 }, { "epoch": 0.98, "learning_rate": 4.1894382562705346e-05, "loss": 0.0011, "step": 6820 }, { "epoch": 0.98, "learning_rate": 4.188436573443385e-05, "loss": 0.0009, "step": 6825 }, { "epoch": 0.99, "learning_rate": 4.187434890616235e-05, "loss": 0.0013, "step": 6830 }, { "epoch": 0.99, "learning_rate": 4.186433207789086e-05, "loss": 0.0013, "step": 6835 }, { "epoch": 0.99, "learning_rate": 4.185431524961936e-05, "loss": 0.0016, "step": 6840 }, { "epoch": 0.99, "learning_rate": 4.184429842134787e-05, "loss": 0.0013, "step": 6845 }, { "epoch": 0.99, "learning_rate": 4.183428159307637e-05, "loss": 0.0008, "step": 6850 }, { "epoch": 0.99, "learning_rate": 4.182426476480488e-05, "loss": 0.0011, "step": 6855 }, { "epoch": 0.99, "learning_rate": 4.181424793653338e-05, "loss": 0.0014, "step": 6860 }, { "epoch": 0.99, "learning_rate": 4.180423110826188e-05, "loss": 0.0013, "step": 6865 }, { "epoch": 0.99, "learning_rate": 4.1794214279990385e-05, "loss": 0.001, "step": 6870 }, { "epoch": 0.99, "learning_rate": 4.178419745171889e-05, "loss": 0.0015, "step": 6875 }, { "epoch": 0.99, "learning_rate": 4.1774180623447395e-05, "loss": 0.0014, "step": 6880 }, { "epoch": 0.99, "learning_rate": 4.17641637951759e-05, "loss": 0.0013, "step": 6885 }, { "epoch": 0.99, "learning_rate": 4.17541469669044e-05, "loss": 0.0009, "step": 6890 }, { "epoch": 0.99, "learning_rate": 4.17441301386329e-05, "loss": 0.0009, "step": 6895 }, { "epoch": 1.0, "learning_rate": 4.173411331036141e-05, "loss": 0.0009, "step": 6900 }, { "epoch": 1.0, "learning_rate": 4.172409648208991e-05, "loss": 0.0013, "step": 6905 }, { "epoch": 1.0, "learning_rate": 4.171407965381841e-05, "loss": 0.001, "step": 6910 }, { "epoch": 1.0, "learning_rate": 4.170406282554692e-05, "loss": 0.0011, "step": 6915 }, { "epoch": 1.0, "learning_rate": 4.169404599727543e-05, "loss": 0.0013, "step": 6920 }, { "epoch": 1.0, "learning_rate": 4.168402916900393e-05, "loss": 0.0014, "step": 6925 }, { "epoch": 1.0, "learning_rate": 4.1674012340732434e-05, "loss": 0.0008, "step": 6930 }, { "epoch": 1.0, "learning_rate": 4.1663995512460936e-05, "loss": 0.0007, "step": 6935 }, { "epoch": 1.0, "learning_rate": 4.165397868418944e-05, "loss": 0.0011, "step": 6940 }, { "epoch": 1.0, "learning_rate": 4.1643961855917946e-05, "loss": 0.001, "step": 6945 }, { "epoch": 1.0, "learning_rate": 4.163394502764645e-05, "loss": 0.001, "step": 6950 }, { "epoch": 1.0, "learning_rate": 4.162392819937495e-05, "loss": 0.0012, "step": 6955 }, { "epoch": 1.0, "learning_rate": 4.161391137110345e-05, "loss": 0.0007, "step": 6960 }, { "epoch": 1.0, "learning_rate": 4.160389454283196e-05, "loss": 0.001, "step": 6965 }, { "epoch": 1.01, "learning_rate": 4.159387771456046e-05, "loss": 0.001, "step": 6970 }, { "epoch": 1.01, "learning_rate": 4.1583860886288964e-05, "loss": 0.001, "step": 6975 }, { "epoch": 1.01, "learning_rate": 4.1573844058017466e-05, "loss": 0.0008, "step": 6980 }, { "epoch": 1.01, "learning_rate": 4.1563827229745975e-05, "loss": 0.0013, "step": 6985 }, { "epoch": 1.01, "learning_rate": 4.155381040147448e-05, "loss": 0.0008, "step": 6990 }, { "epoch": 1.01, "learning_rate": 4.1543793573202985e-05, "loss": 0.0013, "step": 6995 }, { "epoch": 1.01, "learning_rate": 4.153377674493149e-05, "loss": 0.0011, "step": 7000 }, { "epoch": 1.01, "learning_rate": 4.1523759916659996e-05, "loss": 0.0014, "step": 7005 }, { "epoch": 1.01, "learning_rate": 4.15137430883885e-05, "loss": 0.0008, "step": 7010 }, { "epoch": 1.01, "learning_rate": 4.1503726260117e-05, "loss": 0.001, "step": 7015 }, { "epoch": 1.01, "learning_rate": 4.14937094318455e-05, "loss": 0.0008, "step": 7020 }, { "epoch": 1.01, "learning_rate": 4.1483692603574e-05, "loss": 0.0008, "step": 7025 }, { "epoch": 1.01, "learning_rate": 4.147367577530251e-05, "loss": 0.0014, "step": 7030 }, { "epoch": 1.01, "learning_rate": 4.1463658947031013e-05, "loss": 0.0007, "step": 7035 }, { "epoch": 1.02, "learning_rate": 4.1453642118759515e-05, "loss": 0.0012, "step": 7040 }, { "epoch": 1.02, "learning_rate": 4.144362529048802e-05, "loss": 0.0011, "step": 7045 }, { "epoch": 1.02, "learning_rate": 4.1433608462216526e-05, "loss": 0.0012, "step": 7050 }, { "epoch": 1.02, "learning_rate": 4.142359163394503e-05, "loss": 0.0009, "step": 7055 }, { "epoch": 1.02, "learning_rate": 4.1413574805673536e-05, "loss": 0.0012, "step": 7060 }, { "epoch": 1.02, "learning_rate": 4.140355797740204e-05, "loss": 0.0013, "step": 7065 }, { "epoch": 1.02, "learning_rate": 4.139354114913055e-05, "loss": 0.0008, "step": 7070 }, { "epoch": 1.02, "learning_rate": 4.138352432085905e-05, "loss": 0.0008, "step": 7075 }, { "epoch": 1.02, "learning_rate": 4.137350749258755e-05, "loss": 0.0009, "step": 7080 }, { "epoch": 1.02, "learning_rate": 4.136349066431605e-05, "loss": 0.0006, "step": 7085 }, { "epoch": 1.02, "learning_rate": 4.135347383604456e-05, "loss": 0.001, "step": 7090 }, { "epoch": 1.02, "learning_rate": 4.134345700777306e-05, "loss": 0.0011, "step": 7095 }, { "epoch": 1.02, "learning_rate": 4.1333440179501564e-05, "loss": 0.0011, "step": 7100 }, { "epoch": 1.02, "learning_rate": 4.1323423351230066e-05, "loss": 0.0011, "step": 7105 }, { "epoch": 1.03, "learning_rate": 4.131340652295857e-05, "loss": 0.0011, "step": 7110 }, { "epoch": 1.03, "learning_rate": 4.130338969468708e-05, "loss": 0.0014, "step": 7115 }, { "epoch": 1.03, "learning_rate": 4.129337286641558e-05, "loss": 0.0009, "step": 7120 }, { "epoch": 1.03, "learning_rate": 4.128335603814408e-05, "loss": 0.001, "step": 7125 }, { "epoch": 1.03, "learning_rate": 4.127333920987259e-05, "loss": 0.001, "step": 7130 }, { "epoch": 1.03, "learning_rate": 4.126332238160109e-05, "loss": 0.0012, "step": 7135 }, { "epoch": 1.03, "learning_rate": 4.12533055533296e-05, "loss": 0.0009, "step": 7140 }, { "epoch": 1.03, "learning_rate": 4.12432887250581e-05, "loss": 0.0012, "step": 7145 }, { "epoch": 1.03, "learning_rate": 4.12332718967866e-05, "loss": 0.0011, "step": 7150 }, { "epoch": 1.03, "learning_rate": 4.122325506851511e-05, "loss": 0.0007, "step": 7155 }, { "epoch": 1.03, "learning_rate": 4.1213238240243614e-05, "loss": 0.0006, "step": 7160 }, { "epoch": 1.03, "learning_rate": 4.1203221411972115e-05, "loss": 0.0009, "step": 7165 }, { "epoch": 1.03, "learning_rate": 4.119320458370062e-05, "loss": 0.0013, "step": 7170 }, { "epoch": 1.03, "learning_rate": 4.1183187755429126e-05, "loss": 0.0007, "step": 7175 }, { "epoch": 1.04, "learning_rate": 4.117317092715763e-05, "loss": 0.0005, "step": 7180 }, { "epoch": 1.04, "learning_rate": 4.116315409888613e-05, "loss": 0.0011, "step": 7185 }, { "epoch": 1.04, "learning_rate": 4.115313727061463e-05, "loss": 0.0011, "step": 7190 }, { "epoch": 1.04, "learning_rate": 4.114312044234313e-05, "loss": 0.0011, "step": 7195 }, { "epoch": 1.04, "learning_rate": 4.113310361407164e-05, "loss": 0.001, "step": 7200 }, { "epoch": 1.04, "learning_rate": 4.1123086785800144e-05, "loss": 0.0011, "step": 7205 }, { "epoch": 1.04, "learning_rate": 4.111306995752865e-05, "loss": 0.0011, "step": 7210 }, { "epoch": 1.04, "learning_rate": 4.1103053129257154e-05, "loss": 0.0011, "step": 7215 }, { "epoch": 1.04, "learning_rate": 4.109303630098566e-05, "loss": 0.0016, "step": 7220 }, { "epoch": 1.04, "learning_rate": 4.1083019472714165e-05, "loss": 0.0012, "step": 7225 }, { "epoch": 1.04, "learning_rate": 4.1073002644442666e-05, "loss": 0.0008, "step": 7230 }, { "epoch": 1.04, "learning_rate": 4.106298581617117e-05, "loss": 0.0009, "step": 7235 }, { "epoch": 1.04, "learning_rate": 4.105296898789968e-05, "loss": 0.0013, "step": 7240 }, { "epoch": 1.05, "learning_rate": 4.104295215962818e-05, "loss": 0.0014, "step": 7245 }, { "epoch": 1.05, "learning_rate": 4.103293533135668e-05, "loss": 0.0011, "step": 7250 }, { "epoch": 1.05, "learning_rate": 4.102291850308518e-05, "loss": 0.0009, "step": 7255 }, { "epoch": 1.05, "learning_rate": 4.1012901674813684e-05, "loss": 0.0015, "step": 7260 }, { "epoch": 1.05, "learning_rate": 4.100288484654219e-05, "loss": 0.0011, "step": 7265 }, { "epoch": 1.05, "learning_rate": 4.0992868018270695e-05, "loss": 0.001, "step": 7270 }, { "epoch": 1.05, "learning_rate": 4.0982851189999197e-05, "loss": 0.0012, "step": 7275 }, { "epoch": 1.05, "learning_rate": 4.0972834361727705e-05, "loss": 0.0012, "step": 7280 }, { "epoch": 1.05, "learning_rate": 4.096281753345621e-05, "loss": 0.0011, "step": 7285 }, { "epoch": 1.05, "learning_rate": 4.0952800705184716e-05, "loss": 0.0013, "step": 7290 }, { "epoch": 1.05, "learning_rate": 4.094278387691322e-05, "loss": 0.0008, "step": 7295 }, { "epoch": 1.05, "learning_rate": 4.093276704864172e-05, "loss": 0.0009, "step": 7300 }, { "epoch": 1.05, "learning_rate": 4.092275022037023e-05, "loss": 0.001, "step": 7305 }, { "epoch": 1.05, "learning_rate": 4.091273339209873e-05, "loss": 0.0008, "step": 7310 }, { "epoch": 1.06, "learning_rate": 4.090271656382723e-05, "loss": 0.0009, "step": 7315 }, { "epoch": 1.06, "learning_rate": 4.0892699735555733e-05, "loss": 0.0016, "step": 7320 }, { "epoch": 1.06, "learning_rate": 4.088268290728424e-05, "loss": 0.0006, "step": 7325 }, { "epoch": 1.06, "learning_rate": 4.0872666079012744e-05, "loss": 0.0007, "step": 7330 }, { "epoch": 1.06, "learning_rate": 4.0862649250741246e-05, "loss": 0.0004, "step": 7335 }, { "epoch": 1.06, "learning_rate": 4.085263242246975e-05, "loss": 0.0013, "step": 7340 }, { "epoch": 1.06, "learning_rate": 4.084261559419825e-05, "loss": 0.0009, "step": 7345 }, { "epoch": 1.06, "learning_rate": 4.083259876592676e-05, "loss": 0.0007, "step": 7350 }, { "epoch": 1.06, "learning_rate": 4.082258193765526e-05, "loss": 0.0009, "step": 7355 }, { "epoch": 1.06, "learning_rate": 4.081256510938377e-05, "loss": 0.0012, "step": 7360 }, { "epoch": 1.06, "learning_rate": 4.080254828111227e-05, "loss": 0.001, "step": 7365 }, { "epoch": 1.06, "learning_rate": 4.079253145284078e-05, "loss": 0.0008, "step": 7370 }, { "epoch": 1.06, "learning_rate": 4.078251462456928e-05, "loss": 0.001, "step": 7375 }, { "epoch": 1.06, "learning_rate": 4.077249779629778e-05, "loss": 0.001, "step": 7380 }, { "epoch": 1.07, "learning_rate": 4.0762480968026284e-05, "loss": 0.0008, "step": 7385 }, { "epoch": 1.07, "learning_rate": 4.075246413975479e-05, "loss": 0.0009, "step": 7390 }, { "epoch": 1.07, "learning_rate": 4.0742447311483295e-05, "loss": 0.0009, "step": 7395 }, { "epoch": 1.07, "learning_rate": 4.07324304832118e-05, "loss": 0.0009, "step": 7400 }, { "epoch": 1.07, "learning_rate": 4.07224136549403e-05, "loss": 0.0013, "step": 7405 }, { "epoch": 1.07, "learning_rate": 4.071239682666881e-05, "loss": 0.0008, "step": 7410 }, { "epoch": 1.07, "learning_rate": 4.070237999839731e-05, "loss": 0.0013, "step": 7415 }, { "epoch": 1.07, "learning_rate": 4.069236317012581e-05, "loss": 0.0015, "step": 7420 }, { "epoch": 1.07, "learning_rate": 4.068234634185431e-05, "loss": 0.001, "step": 7425 }, { "epoch": 1.07, "learning_rate": 4.067232951358282e-05, "loss": 0.001, "step": 7430 }, { "epoch": 1.07, "learning_rate": 4.066231268531133e-05, "loss": 0.001, "step": 7435 }, { "epoch": 1.07, "learning_rate": 4.065229585703983e-05, "loss": 0.001, "step": 7440 }, { "epoch": 1.07, "learning_rate": 4.0642279028768334e-05, "loss": 0.001, "step": 7445 }, { "epoch": 1.07, "learning_rate": 4.0632262200496835e-05, "loss": 0.0008, "step": 7450 }, { "epoch": 1.08, "learning_rate": 4.0622245372225344e-05, "loss": 0.0012, "step": 7455 }, { "epoch": 1.08, "learning_rate": 4.0612228543953846e-05, "loss": 0.0008, "step": 7460 }, { "epoch": 1.08, "learning_rate": 4.060221171568235e-05, "loss": 0.0007, "step": 7465 }, { "epoch": 1.08, "learning_rate": 4.059219488741085e-05, "loss": 0.001, "step": 7470 }, { "epoch": 1.08, "learning_rate": 4.058217805913936e-05, "loss": 0.0009, "step": 7475 }, { "epoch": 1.08, "learning_rate": 4.057216123086786e-05, "loss": 0.0011, "step": 7480 }, { "epoch": 1.08, "learning_rate": 4.056214440259636e-05, "loss": 0.0011, "step": 7485 }, { "epoch": 1.08, "learning_rate": 4.0552127574324864e-05, "loss": 0.0008, "step": 7490 }, { "epoch": 1.08, "learning_rate": 4.0542110746053366e-05, "loss": 0.0008, "step": 7495 }, { "epoch": 1.08, "learning_rate": 4.0532093917781874e-05, "loss": 0.0015, "step": 7500 }, { "epoch": 1.08, "learning_rate": 4.052207708951038e-05, "loss": 0.0012, "step": 7505 }, { "epoch": 1.08, "learning_rate": 4.0512060261238885e-05, "loss": 0.0014, "step": 7510 }, { "epoch": 1.08, "learning_rate": 4.0502043432967386e-05, "loss": 0.001, "step": 7515 }, { "epoch": 1.08, "learning_rate": 4.0492026604695895e-05, "loss": 0.0012, "step": 7520 }, { "epoch": 1.09, "learning_rate": 4.04820097764244e-05, "loss": 0.0013, "step": 7525 }, { "epoch": 1.09, "learning_rate": 4.04719929481529e-05, "loss": 0.0013, "step": 7530 }, { "epoch": 1.09, "learning_rate": 4.04619761198814e-05, "loss": 0.0011, "step": 7535 }, { "epoch": 1.09, "learning_rate": 4.045195929160991e-05, "loss": 0.0009, "step": 7540 }, { "epoch": 1.09, "learning_rate": 4.044194246333841e-05, "loss": 0.0009, "step": 7545 }, { "epoch": 1.09, "learning_rate": 4.043192563506691e-05, "loss": 0.0007, "step": 7550 }, { "epoch": 1.09, "learning_rate": 4.0421908806795415e-05, "loss": 0.001, "step": 7555 }, { "epoch": 1.09, "learning_rate": 4.041189197852392e-05, "loss": 0.0008, "step": 7560 }, { "epoch": 1.09, "learning_rate": 4.0401875150252425e-05, "loss": 0.0013, "step": 7565 }, { "epoch": 1.09, "learning_rate": 4.039185832198093e-05, "loss": 0.001, "step": 7570 }, { "epoch": 1.09, "learning_rate": 4.0381841493709436e-05, "loss": 0.0012, "step": 7575 }, { "epoch": 1.09, "learning_rate": 4.037182466543794e-05, "loss": 0.001, "step": 7580 }, { "epoch": 1.09, "learning_rate": 4.0361807837166446e-05, "loss": 0.0012, "step": 7585 }, { "epoch": 1.09, "learning_rate": 4.035179100889495e-05, "loss": 0.0013, "step": 7590 }, { "epoch": 1.1, "learning_rate": 4.034177418062345e-05, "loss": 0.0011, "step": 7595 }, { "epoch": 1.1, "learning_rate": 4.033175735235195e-05, "loss": 0.0011, "step": 7600 }, { "epoch": 1.1, "learning_rate": 4.032174052408046e-05, "loss": 0.0012, "step": 7605 }, { "epoch": 1.1, "learning_rate": 4.031172369580896e-05, "loss": 0.0008, "step": 7610 }, { "epoch": 1.1, "learning_rate": 4.0301706867537464e-05, "loss": 0.0012, "step": 7615 }, { "epoch": 1.1, "learning_rate": 4.0291690039265966e-05, "loss": 0.0007, "step": 7620 }, { "epoch": 1.1, "learning_rate": 4.0281673210994474e-05, "loss": 0.0008, "step": 7625 }, { "epoch": 1.1, "learning_rate": 4.0271656382722976e-05, "loss": 0.0009, "step": 7630 }, { "epoch": 1.1, "learning_rate": 4.026163955445148e-05, "loss": 0.0011, "step": 7635 }, { "epoch": 1.1, "learning_rate": 4.025162272617998e-05, "loss": 0.0012, "step": 7640 }, { "epoch": 1.1, "learning_rate": 4.024160589790849e-05, "loss": 0.0011, "step": 7645 }, { "epoch": 1.1, "learning_rate": 4.023158906963699e-05, "loss": 0.001, "step": 7650 }, { "epoch": 1.1, "learning_rate": 4.02215722413655e-05, "loss": 0.0013, "step": 7655 }, { "epoch": 1.1, "learning_rate": 4.0211555413094e-05, "loss": 0.0013, "step": 7660 }, { "epoch": 1.11, "learning_rate": 4.02015385848225e-05, "loss": 0.001, "step": 7665 }, { "epoch": 1.11, "learning_rate": 4.019152175655101e-05, "loss": 0.0012, "step": 7670 }, { "epoch": 1.11, "learning_rate": 4.018150492827951e-05, "loss": 0.0007, "step": 7675 }, { "epoch": 1.11, "learning_rate": 4.0171488100008015e-05, "loss": 0.0011, "step": 7680 }, { "epoch": 1.11, "learning_rate": 4.016147127173652e-05, "loss": 0.0009, "step": 7685 }, { "epoch": 1.11, "learning_rate": 4.0151454443465025e-05, "loss": 0.0011, "step": 7690 }, { "epoch": 1.11, "learning_rate": 4.014143761519353e-05, "loss": 0.0012, "step": 7695 }, { "epoch": 1.11, "learning_rate": 4.013142078692203e-05, "loss": 0.0014, "step": 7700 }, { "epoch": 1.11, "learning_rate": 4.012140395865053e-05, "loss": 0.0009, "step": 7705 }, { "epoch": 1.11, "learning_rate": 4.011138713037904e-05, "loss": 0.0015, "step": 7710 }, { "epoch": 1.11, "learning_rate": 4.010137030210754e-05, "loss": 0.0012, "step": 7715 }, { "epoch": 1.11, "learning_rate": 4.009135347383604e-05, "loss": 0.0009, "step": 7720 }, { "epoch": 1.11, "learning_rate": 4.008133664556455e-05, "loss": 0.001, "step": 7725 }, { "epoch": 1.11, "learning_rate": 4.0071319817293054e-05, "loss": 0.0014, "step": 7730 }, { "epoch": 1.12, "learning_rate": 4.006130298902156e-05, "loss": 0.0013, "step": 7735 }, { "epoch": 1.12, "learning_rate": 4.0051286160750064e-05, "loss": 0.0009, "step": 7740 }, { "epoch": 1.12, "learning_rate": 4.0041269332478566e-05, "loss": 0.0013, "step": 7745 }, { "epoch": 1.12, "learning_rate": 4.003125250420707e-05, "loss": 0.0011, "step": 7750 }, { "epoch": 1.12, "learning_rate": 4.0021235675935576e-05, "loss": 0.0011, "step": 7755 }, { "epoch": 1.12, "learning_rate": 4.001121884766408e-05, "loss": 0.0012, "step": 7760 }, { "epoch": 1.12, "learning_rate": 4.000120201939258e-05, "loss": 0.0009, "step": 7765 }, { "epoch": 1.12, "learning_rate": 3.999118519112108e-05, "loss": 0.0007, "step": 7770 }, { "epoch": 1.12, "learning_rate": 3.998116836284959e-05, "loss": 0.0009, "step": 7775 }, { "epoch": 1.12, "learning_rate": 3.997115153457809e-05, "loss": 0.0008, "step": 7780 }, { "epoch": 1.12, "learning_rate": 3.9961134706306594e-05, "loss": 0.0008, "step": 7785 }, { "epoch": 1.12, "learning_rate": 3.9951117878035096e-05, "loss": 0.0012, "step": 7790 }, { "epoch": 1.12, "learning_rate": 3.9941101049763605e-05, "loss": 0.0009, "step": 7795 }, { "epoch": 1.13, "learning_rate": 3.993108422149211e-05, "loss": 0.001, "step": 7800 }, { "epoch": 1.13, "learning_rate": 3.9921067393220615e-05, "loss": 0.0008, "step": 7805 }, { "epoch": 1.13, "learning_rate": 3.991105056494912e-05, "loss": 0.0014, "step": 7810 }, { "epoch": 1.13, "learning_rate": 3.990103373667762e-05, "loss": 0.0008, "step": 7815 }, { "epoch": 1.13, "learning_rate": 3.989101690840613e-05, "loss": 0.0007, "step": 7820 }, { "epoch": 1.13, "learning_rate": 3.988100008013463e-05, "loss": 0.0011, "step": 7825 }, { "epoch": 1.13, "learning_rate": 3.987098325186313e-05, "loss": 0.0013, "step": 7830 }, { "epoch": 1.13, "learning_rate": 3.986096642359163e-05, "loss": 0.0013, "step": 7835 }, { "epoch": 1.13, "learning_rate": 3.985094959532014e-05, "loss": 0.0011, "step": 7840 }, { "epoch": 1.13, "learning_rate": 3.984093276704864e-05, "loss": 0.0008, "step": 7845 }, { "epoch": 1.13, "learning_rate": 3.9830915938777145e-05, "loss": 0.0007, "step": 7850 }, { "epoch": 1.13, "learning_rate": 3.982089911050565e-05, "loss": 0.0012, "step": 7855 }, { "epoch": 1.13, "learning_rate": 3.9810882282234156e-05, "loss": 0.0008, "step": 7860 }, { "epoch": 1.13, "learning_rate": 3.980086545396266e-05, "loss": 0.001, "step": 7865 }, { "epoch": 1.14, "learning_rate": 3.9790848625691166e-05, "loss": 0.0011, "step": 7870 }, { "epoch": 1.14, "learning_rate": 3.978083179741967e-05, "loss": 0.0008, "step": 7875 }, { "epoch": 1.14, "learning_rate": 3.9770814969148177e-05, "loss": 0.0012, "step": 7880 }, { "epoch": 1.14, "learning_rate": 3.976079814087668e-05, "loss": 0.0012, "step": 7885 }, { "epoch": 1.14, "learning_rate": 3.975078131260518e-05, "loss": 0.0008, "step": 7890 }, { "epoch": 1.14, "learning_rate": 3.974076448433368e-05, "loss": 0.0012, "step": 7895 }, { "epoch": 1.14, "learning_rate": 3.9730747656062184e-05, "loss": 0.0011, "step": 7900 }, { "epoch": 1.14, "learning_rate": 3.972073082779069e-05, "loss": 0.0011, "step": 7905 }, { "epoch": 1.14, "learning_rate": 3.9710713999519194e-05, "loss": 0.0011, "step": 7910 }, { "epoch": 1.14, "learning_rate": 3.9700697171247696e-05, "loss": 0.0016, "step": 7915 }, { "epoch": 1.14, "learning_rate": 3.96906803429762e-05, "loss": 0.0007, "step": 7920 }, { "epoch": 1.14, "learning_rate": 3.968066351470471e-05, "loss": 0.0016, "step": 7925 }, { "epoch": 1.14, "learning_rate": 3.967064668643321e-05, "loss": 0.0008, "step": 7930 }, { "epoch": 1.14, "learning_rate": 3.966062985816171e-05, "loss": 0.001, "step": 7935 }, { "epoch": 1.15, "learning_rate": 3.965061302989022e-05, "loss": 0.0011, "step": 7940 }, { "epoch": 1.15, "learning_rate": 3.964059620161872e-05, "loss": 0.0012, "step": 7945 }, { "epoch": 1.15, "learning_rate": 3.963057937334723e-05, "loss": 0.0009, "step": 7950 }, { "epoch": 1.15, "learning_rate": 3.962056254507573e-05, "loss": 0.001, "step": 7955 }, { "epoch": 1.15, "learning_rate": 3.961054571680423e-05, "loss": 0.0008, "step": 7960 }, { "epoch": 1.15, "learning_rate": 3.9600528888532735e-05, "loss": 0.0005, "step": 7965 }, { "epoch": 1.15, "learning_rate": 3.9590512060261244e-05, "loss": 0.0015, "step": 7970 }, { "epoch": 1.15, "learning_rate": 3.9580495231989745e-05, "loss": 0.0011, "step": 7975 }, { "epoch": 1.15, "learning_rate": 3.957047840371825e-05, "loss": 0.0007, "step": 7980 }, { "epoch": 1.15, "learning_rate": 3.956046157544675e-05, "loss": 0.0007, "step": 7985 }, { "epoch": 1.15, "learning_rate": 3.955044474717526e-05, "loss": 0.0008, "step": 7990 }, { "epoch": 1.15, "learning_rate": 3.954042791890376e-05, "loss": 0.0009, "step": 7995 }, { "epoch": 1.15, "learning_rate": 3.953041109063226e-05, "loss": 0.0009, "step": 8000 }, { "epoch": 1.15, "learning_rate": 3.952039426236076e-05, "loss": 0.0008, "step": 8005 }, { "epoch": 1.16, "learning_rate": 3.951037743408927e-05, "loss": 0.0011, "step": 8010 }, { "epoch": 1.16, "learning_rate": 3.9500360605817774e-05, "loss": 0.0011, "step": 8015 }, { "epoch": 1.16, "learning_rate": 3.949034377754628e-05, "loss": 0.0009, "step": 8020 }, { "epoch": 1.16, "learning_rate": 3.9480326949274784e-05, "loss": 0.0007, "step": 8025 }, { "epoch": 1.16, "learning_rate": 3.947031012100329e-05, "loss": 0.0009, "step": 8030 }, { "epoch": 1.16, "learning_rate": 3.9460293292731795e-05, "loss": 0.0008, "step": 8035 }, { "epoch": 1.16, "learning_rate": 3.9450276464460296e-05, "loss": 0.001, "step": 8040 }, { "epoch": 1.16, "learning_rate": 3.94402596361888e-05, "loss": 0.0011, "step": 8045 }, { "epoch": 1.16, "learning_rate": 3.94302428079173e-05, "loss": 0.0011, "step": 8050 }, { "epoch": 1.16, "learning_rate": 3.942022597964581e-05, "loss": 0.0009, "step": 8055 }, { "epoch": 1.16, "learning_rate": 3.941020915137431e-05, "loss": 0.0011, "step": 8060 }, { "epoch": 1.16, "learning_rate": 3.940019232310281e-05, "loss": 0.0011, "step": 8065 }, { "epoch": 1.16, "learning_rate": 3.9390175494831314e-05, "loss": 0.0008, "step": 8070 }, { "epoch": 1.16, "learning_rate": 3.938015866655982e-05, "loss": 0.0007, "step": 8075 }, { "epoch": 1.17, "learning_rate": 3.9370141838288325e-05, "loss": 0.0008, "step": 8080 }, { "epoch": 1.17, "learning_rate": 3.9360125010016826e-05, "loss": 0.0013, "step": 8085 }, { "epoch": 1.17, "learning_rate": 3.9350108181745335e-05, "loss": 0.001, "step": 8090 }, { "epoch": 1.17, "learning_rate": 3.934009135347384e-05, "loss": 0.0011, "step": 8095 }, { "epoch": 1.17, "learning_rate": 3.9330074525202346e-05, "loss": 0.001, "step": 8100 }, { "epoch": 1.17, "learning_rate": 3.932005769693085e-05, "loss": 0.0012, "step": 8105 }, { "epoch": 1.17, "learning_rate": 3.931004086865935e-05, "loss": 0.0011, "step": 8110 }, { "epoch": 1.17, "learning_rate": 3.930002404038786e-05, "loss": 0.0012, "step": 8115 }, { "epoch": 1.17, "learning_rate": 3.929000721211636e-05, "loss": 0.0009, "step": 8120 }, { "epoch": 1.17, "learning_rate": 3.927999038384486e-05, "loss": 0.0012, "step": 8125 }, { "epoch": 1.17, "learning_rate": 3.926997355557336e-05, "loss": 0.0009, "step": 8130 }, { "epoch": 1.17, "learning_rate": 3.9259956727301865e-05, "loss": 0.0013, "step": 8135 }, { "epoch": 1.17, "learning_rate": 3.9249939899030374e-05, "loss": 0.0009, "step": 8140 }, { "epoch": 1.17, "learning_rate": 3.9239923070758876e-05, "loss": 0.0008, "step": 8145 }, { "epoch": 1.18, "learning_rate": 3.922990624248738e-05, "loss": 0.0011, "step": 8150 }, { "epoch": 1.18, "learning_rate": 3.921988941421588e-05, "loss": 0.0008, "step": 8155 }, { "epoch": 1.18, "learning_rate": 3.920987258594439e-05, "loss": 0.001, "step": 8160 }, { "epoch": 1.18, "learning_rate": 3.919985575767289e-05, "loss": 0.0008, "step": 8165 }, { "epoch": 1.18, "learning_rate": 3.91898389294014e-05, "loss": 0.0008, "step": 8170 }, { "epoch": 1.18, "learning_rate": 3.91798221011299e-05, "loss": 0.0011, "step": 8175 }, { "epoch": 1.18, "learning_rate": 3.916980527285841e-05, "loss": 0.0012, "step": 8180 }, { "epoch": 1.18, "learning_rate": 3.915978844458691e-05, "loss": 0.001, "step": 8185 }, { "epoch": 1.18, "learning_rate": 3.914977161631541e-05, "loss": 0.0008, "step": 8190 }, { "epoch": 1.18, "learning_rate": 3.9139754788043914e-05, "loss": 0.001, "step": 8195 }, { "epoch": 1.18, "learning_rate": 3.912973795977242e-05, "loss": 0.0007, "step": 8200 }, { "epoch": 1.18, "learning_rate": 3.9119721131500925e-05, "loss": 0.0009, "step": 8205 }, { "epoch": 1.18, "learning_rate": 3.910970430322943e-05, "loss": 0.0007, "step": 8210 }, { "epoch": 1.18, "learning_rate": 3.909968747495793e-05, "loss": 0.0009, "step": 8215 }, { "epoch": 1.19, "learning_rate": 3.908967064668643e-05, "loss": 0.0015, "step": 8220 }, { "epoch": 1.19, "learning_rate": 3.907965381841494e-05, "loss": 0.0011, "step": 8225 }, { "epoch": 1.19, "learning_rate": 3.906963699014344e-05, "loss": 0.0009, "step": 8230 }, { "epoch": 1.19, "learning_rate": 3.905962016187194e-05, "loss": 0.0011, "step": 8235 }, { "epoch": 1.19, "learning_rate": 3.904960333360045e-05, "loss": 0.0008, "step": 8240 }, { "epoch": 1.19, "learning_rate": 3.903958650532896e-05, "loss": 0.0009, "step": 8245 }, { "epoch": 1.19, "learning_rate": 3.902956967705746e-05, "loss": 0.0012, "step": 8250 }, { "epoch": 1.19, "learning_rate": 3.9019552848785964e-05, "loss": 0.001, "step": 8255 }, { "epoch": 1.19, "learning_rate": 3.9009536020514465e-05, "loss": 0.0009, "step": 8260 }, { "epoch": 1.19, "learning_rate": 3.8999519192242974e-05, "loss": 0.0012, "step": 8265 }, { "epoch": 1.19, "learning_rate": 3.8989502363971476e-05, "loss": 0.0011, "step": 8270 }, { "epoch": 1.19, "learning_rate": 3.897948553569998e-05, "loss": 0.0011, "step": 8275 }, { "epoch": 1.19, "learning_rate": 3.896946870742848e-05, "loss": 0.0012, "step": 8280 }, { "epoch": 1.2, "learning_rate": 3.895945187915698e-05, "loss": 0.0009, "step": 8285 }, { "epoch": 1.2, "learning_rate": 3.894943505088549e-05, "loss": 0.0009, "step": 8290 }, { "epoch": 1.2, "learning_rate": 3.893941822261399e-05, "loss": 0.0009, "step": 8295 }, { "epoch": 1.2, "learning_rate": 3.8929401394342494e-05, "loss": 0.0007, "step": 8300 }, { "epoch": 1.2, "learning_rate": 3.8919384566070995e-05, "loss": 0.0008, "step": 8305 }, { "epoch": 1.2, "learning_rate": 3.8909367737799504e-05, "loss": 0.0009, "step": 8310 }, { "epoch": 1.2, "learning_rate": 3.889935090952801e-05, "loss": 0.0012, "step": 8315 }, { "epoch": 1.2, "learning_rate": 3.8889334081256515e-05, "loss": 0.0012, "step": 8320 }, { "epoch": 1.2, "learning_rate": 3.8879317252985016e-05, "loss": 0.0009, "step": 8325 }, { "epoch": 1.2, "learning_rate": 3.8869300424713525e-05, "loss": 0.0008, "step": 8330 }, { "epoch": 1.2, "learning_rate": 3.885928359644203e-05, "loss": 0.0011, "step": 8335 }, { "epoch": 1.2, "learning_rate": 3.884926676817053e-05, "loss": 0.0008, "step": 8340 }, { "epoch": 1.2, "learning_rate": 3.883924993989903e-05, "loss": 0.0011, "step": 8345 }, { "epoch": 1.2, "learning_rate": 3.882923311162754e-05, "loss": 0.001, "step": 8350 }, { "epoch": 1.21, "learning_rate": 3.881921628335604e-05, "loss": 0.0013, "step": 8355 }, { "epoch": 1.21, "learning_rate": 3.880919945508454e-05, "loss": 0.0013, "step": 8360 }, { "epoch": 1.21, "learning_rate": 3.8799182626813045e-05, "loss": 0.0007, "step": 8365 }, { "epoch": 1.21, "learning_rate": 3.8789165798541546e-05, "loss": 0.001, "step": 8370 }, { "epoch": 1.21, "learning_rate": 3.8779148970270055e-05, "loss": 0.0008, "step": 8375 }, { "epoch": 1.21, "learning_rate": 3.876913214199856e-05, "loss": 0.0013, "step": 8380 }, { "epoch": 1.21, "learning_rate": 3.8759115313727066e-05, "loss": 0.0012, "step": 8385 }, { "epoch": 1.21, "learning_rate": 3.874909848545557e-05, "loss": 0.0011, "step": 8390 }, { "epoch": 1.21, "learning_rate": 3.8739081657184076e-05, "loss": 0.0006, "step": 8395 }, { "epoch": 1.21, "learning_rate": 3.872906482891258e-05, "loss": 0.0009, "step": 8400 }, { "epoch": 1.21, "learning_rate": 3.871904800064108e-05, "loss": 0.001, "step": 8405 }, { "epoch": 1.21, "learning_rate": 3.870903117236958e-05, "loss": 0.0008, "step": 8410 }, { "epoch": 1.21, "learning_rate": 3.869901434409809e-05, "loss": 0.0013, "step": 8415 }, { "epoch": 1.21, "learning_rate": 3.868899751582659e-05, "loss": 0.0009, "step": 8420 }, { "epoch": 1.22, "learning_rate": 3.8678980687555094e-05, "loss": 0.0013, "step": 8425 }, { "epoch": 1.22, "learning_rate": 3.8668963859283596e-05, "loss": 0.001, "step": 8430 }, { "epoch": 1.22, "learning_rate": 3.8658947031012104e-05, "loss": 0.0007, "step": 8435 }, { "epoch": 1.22, "learning_rate": 3.8648930202740606e-05, "loss": 0.001, "step": 8440 }, { "epoch": 1.22, "learning_rate": 3.863891337446911e-05, "loss": 0.0009, "step": 8445 }, { "epoch": 1.22, "learning_rate": 3.862889654619761e-05, "loss": 0.001, "step": 8450 }, { "epoch": 1.22, "learning_rate": 3.861887971792612e-05, "loss": 0.0014, "step": 8455 }, { "epoch": 1.22, "learning_rate": 3.860886288965462e-05, "loss": 0.0011, "step": 8460 }, { "epoch": 1.22, "learning_rate": 3.859884606138313e-05, "loss": 0.0007, "step": 8465 }, { "epoch": 1.22, "learning_rate": 3.858882923311163e-05, "loss": 0.0008, "step": 8470 }, { "epoch": 1.22, "learning_rate": 3.857881240484013e-05, "loss": 0.0011, "step": 8475 }, { "epoch": 1.22, "learning_rate": 3.856879557656864e-05, "loss": 0.0013, "step": 8480 }, { "epoch": 1.22, "learning_rate": 3.855877874829714e-05, "loss": 0.0007, "step": 8485 }, { "epoch": 1.22, "learning_rate": 3.8548761920025645e-05, "loss": 0.0015, "step": 8490 }, { "epoch": 1.23, "learning_rate": 3.853874509175415e-05, "loss": 0.0012, "step": 8495 }, { "epoch": 1.23, "learning_rate": 3.8528728263482655e-05, "loss": 0.0012, "step": 8500 }, { "epoch": 1.23, "learning_rate": 3.851871143521116e-05, "loss": 0.0013, "step": 8505 }, { "epoch": 1.23, "learning_rate": 3.850869460693966e-05, "loss": 0.0009, "step": 8510 }, { "epoch": 1.23, "learning_rate": 3.849867777866816e-05, "loss": 0.0011, "step": 8515 }, { "epoch": 1.23, "learning_rate": 3.848866095039666e-05, "loss": 0.0007, "step": 8520 }, { "epoch": 1.23, "learning_rate": 3.847864412212517e-05, "loss": 0.0009, "step": 8525 }, { "epoch": 1.23, "learning_rate": 3.846862729385367e-05, "loss": 0.0009, "step": 8530 }, { "epoch": 1.23, "learning_rate": 3.845861046558218e-05, "loss": 0.0009, "step": 8535 }, { "epoch": 1.23, "learning_rate": 3.8448593637310684e-05, "loss": 0.001, "step": 8540 }, { "epoch": 1.23, "learning_rate": 3.843857680903919e-05, "loss": 0.0008, "step": 8545 }, { "epoch": 1.23, "learning_rate": 3.8428559980767694e-05, "loss": 0.001, "step": 8550 }, { "epoch": 1.23, "learning_rate": 3.8418543152496196e-05, "loss": 0.001, "step": 8555 }, { "epoch": 1.23, "learning_rate": 3.84085263242247e-05, "loss": 0.0011, "step": 8560 }, { "epoch": 1.24, "learning_rate": 3.8398509495953206e-05, "loss": 0.001, "step": 8565 }, { "epoch": 1.24, "learning_rate": 3.838849266768171e-05, "loss": 0.0011, "step": 8570 }, { "epoch": 1.24, "learning_rate": 3.837847583941021e-05, "loss": 0.0012, "step": 8575 }, { "epoch": 1.24, "learning_rate": 3.836845901113871e-05, "loss": 0.0012, "step": 8580 }, { "epoch": 1.24, "learning_rate": 3.835844218286722e-05, "loss": 0.0006, "step": 8585 }, { "epoch": 1.24, "learning_rate": 3.834842535459572e-05, "loss": 0.0012, "step": 8590 }, { "epoch": 1.24, "learning_rate": 3.8338408526324224e-05, "loss": 0.0009, "step": 8595 }, { "epoch": 1.24, "learning_rate": 3.8328391698052726e-05, "loss": 0.0013, "step": 8600 }, { "epoch": 1.24, "learning_rate": 3.8318374869781235e-05, "loss": 0.001, "step": 8605 }, { "epoch": 1.24, "learning_rate": 3.830835804150974e-05, "loss": 0.0011, "step": 8610 }, { "epoch": 1.24, "learning_rate": 3.8298341213238245e-05, "loss": 0.0011, "step": 8615 }, { "epoch": 1.24, "learning_rate": 3.828832438496675e-05, "loss": 0.0013, "step": 8620 }, { "epoch": 1.24, "learning_rate": 3.827830755669525e-05, "loss": 0.0013, "step": 8625 }, { "epoch": 1.24, "learning_rate": 3.826829072842376e-05, "loss": 0.0008, "step": 8630 }, { "epoch": 1.25, "learning_rate": 3.825827390015226e-05, "loss": 0.0009, "step": 8635 }, { "epoch": 1.25, "learning_rate": 3.824825707188076e-05, "loss": 0.0011, "step": 8640 }, { "epoch": 1.25, "learning_rate": 3.823824024360926e-05, "loss": 0.0011, "step": 8645 }, { "epoch": 1.25, "learning_rate": 3.822822341533777e-05, "loss": 0.001, "step": 8650 }, { "epoch": 1.25, "learning_rate": 3.821820658706627e-05, "loss": 0.0011, "step": 8655 }, { "epoch": 1.25, "learning_rate": 3.8208189758794775e-05, "loss": 0.0009, "step": 8660 }, { "epoch": 1.25, "learning_rate": 3.819817293052328e-05, "loss": 0.0006, "step": 8665 }, { "epoch": 1.25, "learning_rate": 3.8188156102251786e-05, "loss": 0.001, "step": 8670 }, { "epoch": 1.25, "learning_rate": 3.817813927398029e-05, "loss": 0.0009, "step": 8675 }, { "epoch": 1.25, "learning_rate": 3.8168122445708796e-05, "loss": 0.0012, "step": 8680 }, { "epoch": 1.25, "learning_rate": 3.81581056174373e-05, "loss": 0.0013, "step": 8685 }, { "epoch": 1.25, "learning_rate": 3.81480887891658e-05, "loss": 0.0009, "step": 8690 }, { "epoch": 1.25, "learning_rate": 3.813807196089431e-05, "loss": 0.0011, "step": 8695 }, { "epoch": 1.25, "learning_rate": 3.812805513262281e-05, "loss": 0.0011, "step": 8700 }, { "epoch": 1.26, "learning_rate": 3.811803830435131e-05, "loss": 0.0009, "step": 8705 }, { "epoch": 1.26, "learning_rate": 3.8108021476079814e-05, "loss": 0.0011, "step": 8710 }, { "epoch": 1.26, "learning_rate": 3.809800464780832e-05, "loss": 0.0011, "step": 8715 }, { "epoch": 1.26, "learning_rate": 3.8087987819536824e-05, "loss": 0.001, "step": 8720 }, { "epoch": 1.26, "learning_rate": 3.8077970991265326e-05, "loss": 0.0007, "step": 8725 }, { "epoch": 1.26, "learning_rate": 3.806795416299383e-05, "loss": 0.0011, "step": 8730 }, { "epoch": 1.26, "learning_rate": 3.8057937334722337e-05, "loss": 0.0005, "step": 8735 }, { "epoch": 1.26, "learning_rate": 3.804792050645084e-05, "loss": 0.001, "step": 8740 }, { "epoch": 1.26, "learning_rate": 3.803790367817934e-05, "loss": 0.0007, "step": 8745 }, { "epoch": 1.26, "learning_rate": 3.802788684990785e-05, "loss": 0.0011, "step": 8750 }, { "epoch": 1.26, "learning_rate": 3.801787002163635e-05, "loss": 0.001, "step": 8755 }, { "epoch": 1.26, "learning_rate": 3.800785319336486e-05, "loss": 0.0007, "step": 8760 }, { "epoch": 1.26, "learning_rate": 3.799783636509336e-05, "loss": 0.0009, "step": 8765 }, { "epoch": 1.26, "learning_rate": 3.798781953682186e-05, "loss": 0.0008, "step": 8770 }, { "epoch": 1.27, "learning_rate": 3.7977802708550365e-05, "loss": 0.0009, "step": 8775 }, { "epoch": 1.27, "learning_rate": 3.7967785880278873e-05, "loss": 0.001, "step": 8780 }, { "epoch": 1.27, "learning_rate": 3.7957769052007375e-05, "loss": 0.0008, "step": 8785 }, { "epoch": 1.27, "learning_rate": 3.794775222373588e-05, "loss": 0.001, "step": 8790 }, { "epoch": 1.27, "learning_rate": 3.793773539546438e-05, "loss": 0.0008, "step": 8795 }, { "epoch": 1.27, "learning_rate": 3.792771856719289e-05, "loss": 0.001, "step": 8800 }, { "epoch": 1.27, "learning_rate": 3.791770173892139e-05, "loss": 0.0014, "step": 8805 }, { "epoch": 1.27, "learning_rate": 3.790768491064989e-05, "loss": 0.0012, "step": 8810 }, { "epoch": 1.27, "learning_rate": 3.789766808237839e-05, "loss": 0.0008, "step": 8815 }, { "epoch": 1.27, "learning_rate": 3.78876512541069e-05, "loss": 0.0012, "step": 8820 }, { "epoch": 1.27, "learning_rate": 3.7877634425835404e-05, "loss": 0.0011, "step": 8825 }, { "epoch": 1.27, "learning_rate": 3.786761759756391e-05, "loss": 0.0012, "step": 8830 }, { "epoch": 1.27, "learning_rate": 3.7857600769292414e-05, "loss": 0.0011, "step": 8835 }, { "epoch": 1.28, "learning_rate": 3.7847583941020916e-05, "loss": 0.0009, "step": 8840 }, { "epoch": 1.28, "learning_rate": 3.7837567112749424e-05, "loss": 0.0006, "step": 8845 }, { "epoch": 1.28, "learning_rate": 3.7827550284477926e-05, "loss": 0.0009, "step": 8850 }, { "epoch": 1.28, "learning_rate": 3.781753345620643e-05, "loss": 0.0011, "step": 8855 }, { "epoch": 1.28, "learning_rate": 3.780751662793493e-05, "loss": 0.0008, "step": 8860 }, { "epoch": 1.28, "learning_rate": 3.779749979966344e-05, "loss": 0.001, "step": 8865 }, { "epoch": 1.28, "learning_rate": 3.778748297139194e-05, "loss": 0.0011, "step": 8870 }, { "epoch": 1.28, "learning_rate": 3.777746614312044e-05, "loss": 0.0014, "step": 8875 }, { "epoch": 1.28, "learning_rate": 3.7767449314848944e-05, "loss": 0.0009, "step": 8880 }, { "epoch": 1.28, "learning_rate": 3.775743248657745e-05, "loss": 0.0012, "step": 8885 }, { "epoch": 1.28, "learning_rate": 3.7747415658305955e-05, "loss": 0.0009, "step": 8890 }, { "epoch": 1.28, "learning_rate": 3.7737398830034456e-05, "loss": 0.0011, "step": 8895 }, { "epoch": 1.28, "learning_rate": 3.7727382001762965e-05, "loss": 0.0009, "step": 8900 }, { "epoch": 1.28, "learning_rate": 3.771736517349147e-05, "loss": 0.0011, "step": 8905 }, { "epoch": 1.29, "learning_rate": 3.7707348345219975e-05, "loss": 0.0008, "step": 8910 }, { "epoch": 1.29, "learning_rate": 3.769733151694848e-05, "loss": 0.001, "step": 8915 }, { "epoch": 1.29, "learning_rate": 3.768731468867698e-05, "loss": 0.0008, "step": 8920 }, { "epoch": 1.29, "learning_rate": 3.767729786040548e-05, "loss": 0.001, "step": 8925 }, { "epoch": 1.29, "learning_rate": 3.766728103213399e-05, "loss": 0.0009, "step": 8930 }, { "epoch": 1.29, "learning_rate": 3.765726420386249e-05, "loss": 0.0006, "step": 8935 }, { "epoch": 1.29, "learning_rate": 3.764724737559099e-05, "loss": 0.001, "step": 8940 }, { "epoch": 1.29, "learning_rate": 3.7637230547319495e-05, "loss": 0.001, "step": 8945 }, { "epoch": 1.29, "learning_rate": 3.7627213719048004e-05, "loss": 0.0011, "step": 8950 }, { "epoch": 1.29, "learning_rate": 3.7617196890776506e-05, "loss": 0.0008, "step": 8955 }, { "epoch": 1.29, "learning_rate": 3.760718006250501e-05, "loss": 0.0008, "step": 8960 }, { "epoch": 1.29, "learning_rate": 3.759716323423351e-05, "loss": 0.001, "step": 8965 }, { "epoch": 1.29, "learning_rate": 3.758714640596202e-05, "loss": 0.0011, "step": 8970 }, { "epoch": 1.29, "learning_rate": 3.757712957769052e-05, "loss": 0.0009, "step": 8975 }, { "epoch": 1.3, "learning_rate": 3.756711274941903e-05, "loss": 0.0007, "step": 8980 }, { "epoch": 1.3, "learning_rate": 3.755709592114753e-05, "loss": 0.0009, "step": 8985 }, { "epoch": 1.3, "learning_rate": 3.754707909287603e-05, "loss": 0.0008, "step": 8990 }, { "epoch": 1.3, "learning_rate": 3.753706226460454e-05, "loss": 0.0007, "step": 8995 }, { "epoch": 1.3, "learning_rate": 3.752704543633304e-05, "loss": 0.001, "step": 9000 }, { "epoch": 1.3, "learning_rate": 3.7517028608061544e-05, "loss": 0.0012, "step": 9005 }, { "epoch": 1.3, "learning_rate": 3.7507011779790046e-05, "loss": 0.0015, "step": 9010 }, { "epoch": 1.3, "learning_rate": 3.7496994951518555e-05, "loss": 0.0009, "step": 9015 }, { "epoch": 1.3, "learning_rate": 3.7486978123247057e-05, "loss": 0.0007, "step": 9020 }, { "epoch": 1.3, "learning_rate": 3.747696129497556e-05, "loss": 0.0008, "step": 9025 }, { "epoch": 1.3, "learning_rate": 3.746694446670406e-05, "loss": 0.0006, "step": 9030 }, { "epoch": 1.3, "learning_rate": 3.745692763843257e-05, "loss": 0.0012, "step": 9035 }, { "epoch": 1.3, "learning_rate": 3.744691081016107e-05, "loss": 0.0011, "step": 9040 }, { "epoch": 1.3, "learning_rate": 3.743689398188957e-05, "loss": 0.0007, "step": 9045 }, { "epoch": 1.31, "learning_rate": 3.742687715361808e-05, "loss": 0.0012, "step": 9050 }, { "epoch": 1.31, "learning_rate": 3.741686032534659e-05, "loss": 0.0008, "step": 9055 }, { "epoch": 1.31, "learning_rate": 3.740684349707509e-05, "loss": 0.001, "step": 9060 }, { "epoch": 1.31, "learning_rate": 3.7396826668803593e-05, "loss": 0.0011, "step": 9065 }, { "epoch": 1.31, "learning_rate": 3.7386809840532095e-05, "loss": 0.0008, "step": 9070 }, { "epoch": 1.31, "learning_rate": 3.73767930122606e-05, "loss": 0.0009, "step": 9075 }, { "epoch": 1.31, "learning_rate": 3.7366776183989106e-05, "loss": 0.001, "step": 9080 }, { "epoch": 1.31, "learning_rate": 3.735675935571761e-05, "loss": 0.0009, "step": 9085 }, { "epoch": 1.31, "learning_rate": 3.734674252744611e-05, "loss": 0.0008, "step": 9090 }, { "epoch": 1.31, "learning_rate": 3.733672569917461e-05, "loss": 0.0004, "step": 9095 }, { "epoch": 1.31, "learning_rate": 3.732670887090312e-05, "loss": 0.0009, "step": 9100 }, { "epoch": 1.31, "learning_rate": 3.731669204263162e-05, "loss": 0.0012, "step": 9105 }, { "epoch": 1.31, "learning_rate": 3.7306675214360124e-05, "loss": 0.001, "step": 9110 }, { "epoch": 1.31, "learning_rate": 3.7296658386088625e-05, "loss": 0.0007, "step": 9115 }, { "epoch": 1.32, "learning_rate": 3.7286641557817134e-05, "loss": 0.0012, "step": 9120 }, { "epoch": 1.32, "learning_rate": 3.727662472954564e-05, "loss": 0.001, "step": 9125 }, { "epoch": 1.32, "learning_rate": 3.7266607901274144e-05, "loss": 0.0011, "step": 9130 }, { "epoch": 1.32, "learning_rate": 3.7256591073002646e-05, "loss": 0.001, "step": 9135 }, { "epoch": 1.32, "learning_rate": 3.7246574244731155e-05, "loss": 0.0007, "step": 9140 }, { "epoch": 1.32, "learning_rate": 3.723655741645966e-05, "loss": 0.0008, "step": 9145 }, { "epoch": 1.32, "learning_rate": 3.722654058818816e-05, "loss": 0.0006, "step": 9150 }, { "epoch": 1.32, "learning_rate": 3.721652375991666e-05, "loss": 0.001, "step": 9155 }, { "epoch": 1.32, "learning_rate": 3.720650693164516e-05, "loss": 0.0008, "step": 9160 }, { "epoch": 1.32, "learning_rate": 3.719649010337367e-05, "loss": 0.001, "step": 9165 }, { "epoch": 1.32, "learning_rate": 3.718647327510217e-05, "loss": 0.0009, "step": 9170 }, { "epoch": 1.32, "learning_rate": 3.7176456446830675e-05, "loss": 0.0005, "step": 9175 }, { "epoch": 1.32, "learning_rate": 3.7166439618559176e-05, "loss": 0.0008, "step": 9180 }, { "epoch": 1.32, "learning_rate": 3.7156422790287685e-05, "loss": 0.0011, "step": 9185 }, { "epoch": 1.33, "learning_rate": 3.714640596201619e-05, "loss": 0.0008, "step": 9190 }, { "epoch": 1.33, "learning_rate": 3.7136389133744695e-05, "loss": 0.0008, "step": 9195 }, { "epoch": 1.33, "learning_rate": 3.71263723054732e-05, "loss": 0.0005, "step": 9200 }, { "epoch": 1.33, "learning_rate": 3.7116355477201706e-05, "loss": 0.0011, "step": 9205 }, { "epoch": 1.33, "learning_rate": 3.710633864893021e-05, "loss": 0.0011, "step": 9210 }, { "epoch": 1.33, "learning_rate": 3.709632182065871e-05, "loss": 0.0009, "step": 9215 }, { "epoch": 1.33, "learning_rate": 3.708630499238721e-05, "loss": 0.0009, "step": 9220 }, { "epoch": 1.33, "learning_rate": 3.707628816411572e-05, "loss": 0.001, "step": 9225 }, { "epoch": 1.33, "learning_rate": 3.706627133584422e-05, "loss": 0.0007, "step": 9230 }, { "epoch": 1.33, "learning_rate": 3.7056254507572724e-05, "loss": 0.0005, "step": 9235 }, { "epoch": 1.33, "learning_rate": 3.7046237679301226e-05, "loss": 0.0013, "step": 9240 }, { "epoch": 1.33, "learning_rate": 3.703622085102973e-05, "loss": 0.0011, "step": 9245 }, { "epoch": 1.33, "learning_rate": 3.7026204022758236e-05, "loss": 0.0012, "step": 9250 }, { "epoch": 1.33, "learning_rate": 3.701618719448674e-05, "loss": 0.0008, "step": 9255 }, { "epoch": 1.34, "learning_rate": 3.700617036621524e-05, "loss": 0.0007, "step": 9260 }, { "epoch": 1.34, "learning_rate": 3.699615353794375e-05, "loss": 0.0006, "step": 9265 }, { "epoch": 1.34, "learning_rate": 3.698613670967225e-05, "loss": 0.0009, "step": 9270 }, { "epoch": 1.34, "learning_rate": 3.697611988140076e-05, "loss": 0.0007, "step": 9275 }, { "epoch": 1.34, "learning_rate": 3.696610305312926e-05, "loss": 0.0009, "step": 9280 }, { "epoch": 1.34, "learning_rate": 3.695608622485776e-05, "loss": 0.0008, "step": 9285 }, { "epoch": 1.34, "learning_rate": 3.694606939658627e-05, "loss": 0.0008, "step": 9290 }, { "epoch": 1.34, "learning_rate": 3.693605256831477e-05, "loss": 0.0009, "step": 9295 }, { "epoch": 1.34, "learning_rate": 3.6926035740043275e-05, "loss": 0.0014, "step": 9300 }, { "epoch": 1.34, "learning_rate": 3.6916018911771777e-05, "loss": 0.0012, "step": 9305 }, { "epoch": 1.34, "learning_rate": 3.690600208350028e-05, "loss": 0.0007, "step": 9310 }, { "epoch": 1.34, "learning_rate": 3.689598525522879e-05, "loss": 0.0012, "step": 9315 }, { "epoch": 1.34, "learning_rate": 3.688596842695729e-05, "loss": 0.001, "step": 9320 }, { "epoch": 1.35, "learning_rate": 3.687595159868579e-05, "loss": 0.001, "step": 9325 }, { "epoch": 1.35, "learning_rate": 3.686593477041429e-05, "loss": 0.0011, "step": 9330 }, { "epoch": 1.35, "learning_rate": 3.68559179421428e-05, "loss": 0.0009, "step": 9335 }, { "epoch": 1.35, "learning_rate": 3.68459011138713e-05, "loss": 0.0011, "step": 9340 }, { "epoch": 1.35, "learning_rate": 3.683588428559981e-05, "loss": 0.001, "step": 9345 }, { "epoch": 1.35, "learning_rate": 3.6825867457328313e-05, "loss": 0.0006, "step": 9350 }, { "epoch": 1.35, "learning_rate": 3.681585062905682e-05, "loss": 0.0009, "step": 9355 }, { "epoch": 1.35, "learning_rate": 3.6805833800785324e-05, "loss": 0.0008, "step": 9360 }, { "epoch": 1.35, "learning_rate": 3.6795816972513826e-05, "loss": 0.0007, "step": 9365 }, { "epoch": 1.35, "learning_rate": 3.678580014424233e-05, "loss": 0.0008, "step": 9370 }, { "epoch": 1.35, "learning_rate": 3.6775783315970836e-05, "loss": 0.0013, "step": 9375 }, { "epoch": 1.35, "learning_rate": 3.676576648769934e-05, "loss": 0.0013, "step": 9380 }, { "epoch": 1.35, "learning_rate": 3.675574965942784e-05, "loss": 0.0009, "step": 9385 }, { "epoch": 1.35, "learning_rate": 3.674573283115634e-05, "loss": 0.0012, "step": 9390 }, { "epoch": 1.36, "learning_rate": 3.6735716002884844e-05, "loss": 0.0009, "step": 9395 }, { "epoch": 1.36, "learning_rate": 3.672569917461335e-05, "loss": 0.001, "step": 9400 }, { "epoch": 1.36, "learning_rate": 3.6715682346341854e-05, "loss": 0.001, "step": 9405 }, { "epoch": 1.36, "learning_rate": 3.6705665518070356e-05, "loss": 0.001, "step": 9410 }, { "epoch": 1.36, "learning_rate": 3.6695648689798864e-05, "loss": 0.0009, "step": 9415 }, { "epoch": 1.36, "learning_rate": 3.668563186152737e-05, "loss": 0.0013, "step": 9420 }, { "epoch": 1.36, "learning_rate": 3.6675615033255875e-05, "loss": 0.0012, "step": 9425 }, { "epoch": 1.36, "learning_rate": 3.666559820498438e-05, "loss": 0.0009, "step": 9430 }, { "epoch": 1.36, "learning_rate": 3.665558137671288e-05, "loss": 0.0008, "step": 9435 }, { "epoch": 1.36, "learning_rate": 3.664556454844139e-05, "loss": 0.0011, "step": 9440 }, { "epoch": 1.36, "learning_rate": 3.663554772016989e-05, "loss": 0.0013, "step": 9445 }, { "epoch": 1.36, "learning_rate": 3.662553089189839e-05, "loss": 0.0009, "step": 9450 }, { "epoch": 1.36, "learning_rate": 3.661551406362689e-05, "loss": 0.0008, "step": 9455 }, { "epoch": 1.36, "learning_rate": 3.66054972353554e-05, "loss": 0.0011, "step": 9460 }, { "epoch": 1.37, "learning_rate": 3.65954804070839e-05, "loss": 0.0012, "step": 9465 }, { "epoch": 1.37, "learning_rate": 3.6585463578812405e-05, "loss": 0.001, "step": 9470 }, { "epoch": 1.37, "learning_rate": 3.657544675054091e-05, "loss": 0.0012, "step": 9475 }, { "epoch": 1.37, "learning_rate": 3.656542992226941e-05, "loss": 0.0007, "step": 9480 }, { "epoch": 1.37, "learning_rate": 3.655541309399792e-05, "loss": 0.0008, "step": 9485 }, { "epoch": 1.37, "learning_rate": 3.6545396265726426e-05, "loss": 0.0009, "step": 9490 }, { "epoch": 1.37, "learning_rate": 3.653537943745493e-05, "loss": 0.0008, "step": 9495 }, { "epoch": 1.37, "learning_rate": 3.652536260918343e-05, "loss": 0.0012, "step": 9500 }, { "epoch": 1.37, "learning_rate": 3.651534578091194e-05, "loss": 0.0009, "step": 9505 }, { "epoch": 1.37, "learning_rate": 3.650532895264044e-05, "loss": 0.0005, "step": 9510 }, { "epoch": 1.37, "learning_rate": 3.649531212436894e-05, "loss": 0.0007, "step": 9515 }, { "epoch": 1.37, "learning_rate": 3.6485295296097444e-05, "loss": 0.0006, "step": 9520 }, { "epoch": 1.37, "learning_rate": 3.647527846782595e-05, "loss": 0.0007, "step": 9525 }, { "epoch": 1.37, "learning_rate": 3.6465261639554454e-05, "loss": 0.0007, "step": 9530 }, { "epoch": 1.38, "learning_rate": 3.6455244811282956e-05, "loss": 0.001, "step": 9535 }, { "epoch": 1.38, "learning_rate": 3.644522798301146e-05, "loss": 0.0012, "step": 9540 }, { "epoch": 1.38, "learning_rate": 3.643521115473996e-05, "loss": 0.0009, "step": 9545 }, { "epoch": 1.38, "learning_rate": 3.642519432646847e-05, "loss": 0.001, "step": 9550 }, { "epoch": 1.38, "learning_rate": 3.641517749819697e-05, "loss": 0.0007, "step": 9555 }, { "epoch": 1.38, "learning_rate": 3.640516066992548e-05, "loss": 0.0007, "step": 9560 }, { "epoch": 1.38, "learning_rate": 3.639514384165398e-05, "loss": 0.0008, "step": 9565 }, { "epoch": 1.38, "learning_rate": 3.638512701338249e-05, "loss": 0.0005, "step": 9570 }, { "epoch": 1.38, "learning_rate": 3.637511018511099e-05, "loss": 0.0008, "step": 9575 }, { "epoch": 1.38, "learning_rate": 3.636509335683949e-05, "loss": 0.0008, "step": 9580 }, { "epoch": 1.38, "learning_rate": 3.6355076528567995e-05, "loss": 0.0012, "step": 9585 }, { "epoch": 1.38, "learning_rate": 3.63450597002965e-05, "loss": 0.0008, "step": 9590 }, { "epoch": 1.38, "learning_rate": 3.6335042872025005e-05, "loss": 0.0009, "step": 9595 }, { "epoch": 1.38, "learning_rate": 3.632502604375351e-05, "loss": 0.001, "step": 9600 }, { "epoch": 1.39, "learning_rate": 3.631500921548201e-05, "loss": 0.0016, "step": 9605 }, { "epoch": 1.39, "learning_rate": 3.630499238721052e-05, "loss": 0.0008, "step": 9610 }, { "epoch": 1.39, "learning_rate": 3.629497555893902e-05, "loss": 0.0013, "step": 9615 }, { "epoch": 1.39, "learning_rate": 3.628495873066752e-05, "loss": 0.001, "step": 9620 }, { "epoch": 1.39, "learning_rate": 3.627494190239602e-05, "loss": 0.0007, "step": 9625 }, { "epoch": 1.39, "learning_rate": 3.626492507412453e-05, "loss": 0.0008, "step": 9630 }, { "epoch": 1.39, "learning_rate": 3.6254908245853033e-05, "loss": 0.0008, "step": 9635 }, { "epoch": 1.39, "learning_rate": 3.624489141758154e-05, "loss": 0.0014, "step": 9640 }, { "epoch": 1.39, "learning_rate": 3.6234874589310044e-05, "loss": 0.0009, "step": 9645 }, { "epoch": 1.39, "learning_rate": 3.6224857761038546e-05, "loss": 0.0008, "step": 9650 }, { "epoch": 1.39, "learning_rate": 3.6214840932767054e-05, "loss": 0.0008, "step": 9655 }, { "epoch": 1.39, "learning_rate": 3.6204824104495556e-05, "loss": 0.0011, "step": 9660 }, { "epoch": 1.39, "learning_rate": 3.619480727622406e-05, "loss": 0.0008, "step": 9665 }, { "epoch": 1.39, "learning_rate": 3.618479044795256e-05, "loss": 0.0014, "step": 9670 }, { "epoch": 1.4, "learning_rate": 3.617477361968107e-05, "loss": 0.0008, "step": 9675 }, { "epoch": 1.4, "learning_rate": 3.616475679140957e-05, "loss": 0.001, "step": 9680 }, { "epoch": 1.4, "learning_rate": 3.615473996313807e-05, "loss": 0.001, "step": 9685 }, { "epoch": 1.4, "learning_rate": 3.6144723134866574e-05, "loss": 0.0008, "step": 9690 }, { "epoch": 1.4, "learning_rate": 3.613470630659508e-05, "loss": 0.001, "step": 9695 }, { "epoch": 1.4, "learning_rate": 3.6124689478323584e-05, "loss": 0.0012, "step": 9700 }, { "epoch": 1.4, "learning_rate": 3.6114672650052086e-05, "loss": 0.0009, "step": 9705 }, { "epoch": 1.4, "learning_rate": 3.6104655821780595e-05, "loss": 0.0009, "step": 9710 }, { "epoch": 1.4, "learning_rate": 3.60946389935091e-05, "loss": 0.0006, "step": 9715 }, { "epoch": 1.4, "learning_rate": 3.6084622165237605e-05, "loss": 0.001, "step": 9720 }, { "epoch": 1.4, "learning_rate": 3.607460533696611e-05, "loss": 0.0007, "step": 9725 }, { "epoch": 1.4, "learning_rate": 3.606458850869461e-05, "loss": 0.001, "step": 9730 }, { "epoch": 1.4, "learning_rate": 3.605457168042311e-05, "loss": 0.001, "step": 9735 }, { "epoch": 1.4, "learning_rate": 3.604455485215162e-05, "loss": 0.0007, "step": 9740 }, { "epoch": 1.41, "learning_rate": 3.603453802388012e-05, "loss": 0.0011, "step": 9745 }, { "epoch": 1.41, "learning_rate": 3.602452119560862e-05, "loss": 0.001, "step": 9750 }, { "epoch": 1.41, "learning_rate": 3.6014504367337125e-05, "loss": 0.0008, "step": 9755 }, { "epoch": 1.41, "learning_rate": 3.6004487539065634e-05, "loss": 0.0007, "step": 9760 }, { "epoch": 1.41, "learning_rate": 3.5994470710794135e-05, "loss": 0.0012, "step": 9765 }, { "epoch": 1.41, "learning_rate": 3.598445388252264e-05, "loss": 0.0008, "step": 9770 }, { "epoch": 1.41, "learning_rate": 3.597443705425114e-05, "loss": 0.001, "step": 9775 }, { "epoch": 1.41, "learning_rate": 3.596442022597965e-05, "loss": 0.0009, "step": 9780 }, { "epoch": 1.41, "learning_rate": 3.595440339770815e-05, "loss": 0.001, "step": 9785 }, { "epoch": 1.41, "learning_rate": 3.594438656943666e-05, "loss": 0.0009, "step": 9790 }, { "epoch": 1.41, "learning_rate": 3.593436974116516e-05, "loss": 0.0013, "step": 9795 }, { "epoch": 1.41, "learning_rate": 3.592435291289366e-05, "loss": 0.0007, "step": 9800 }, { "epoch": 1.41, "learning_rate": 3.591433608462217e-05, "loss": 0.0011, "step": 9805 }, { "epoch": 1.41, "learning_rate": 3.590431925635067e-05, "loss": 0.0011, "step": 9810 }, { "epoch": 1.42, "learning_rate": 3.5894302428079174e-05, "loss": 0.0007, "step": 9815 }, { "epoch": 1.42, "learning_rate": 3.5884285599807676e-05, "loss": 0.0007, "step": 9820 }, { "epoch": 1.42, "learning_rate": 3.5874268771536185e-05, "loss": 0.0011, "step": 9825 }, { "epoch": 1.42, "learning_rate": 3.5864251943264686e-05, "loss": 0.001, "step": 9830 }, { "epoch": 1.42, "learning_rate": 3.585423511499319e-05, "loss": 0.0008, "step": 9835 }, { "epoch": 1.42, "learning_rate": 3.584421828672169e-05, "loss": 0.0008, "step": 9840 }, { "epoch": 1.42, "learning_rate": 3.58342014584502e-05, "loss": 0.0009, "step": 9845 }, { "epoch": 1.42, "learning_rate": 3.58241846301787e-05, "loss": 0.0006, "step": 9850 }, { "epoch": 1.42, "learning_rate": 3.58141678019072e-05, "loss": 0.0007, "step": 9855 }, { "epoch": 1.42, "learning_rate": 3.580415097363571e-05, "loss": 0.0007, "step": 9860 }, { "epoch": 1.42, "learning_rate": 3.579413414536421e-05, "loss": 0.001, "step": 9865 }, { "epoch": 1.42, "learning_rate": 3.578411731709272e-05, "loss": 0.0008, "step": 9870 }, { "epoch": 1.42, "learning_rate": 3.577410048882122e-05, "loss": 0.001, "step": 9875 }, { "epoch": 1.43, "learning_rate": 3.5764083660549725e-05, "loss": 0.0009, "step": 9880 }, { "epoch": 1.43, "learning_rate": 3.575406683227823e-05, "loss": 0.0007, "step": 9885 }, { "epoch": 1.43, "learning_rate": 3.5744050004006736e-05, "loss": 0.0008, "step": 9890 }, { "epoch": 1.43, "learning_rate": 3.573403317573524e-05, "loss": 0.0012, "step": 9895 }, { "epoch": 1.43, "learning_rate": 3.572401634746374e-05, "loss": 0.0012, "step": 9900 }, { "epoch": 1.43, "learning_rate": 3.571399951919224e-05, "loss": 0.0009, "step": 9905 }, { "epoch": 1.43, "learning_rate": 3.570398269092075e-05, "loss": 0.0011, "step": 9910 }, { "epoch": 1.43, "learning_rate": 3.569396586264925e-05, "loss": 0.0009, "step": 9915 }, { "epoch": 1.43, "learning_rate": 3.5683949034377753e-05, "loss": 0.0012, "step": 9920 }, { "epoch": 1.43, "learning_rate": 3.5673932206106255e-05, "loss": 0.001, "step": 9925 }, { "epoch": 1.43, "learning_rate": 3.5663915377834764e-05, "loss": 0.001, "step": 9930 }, { "epoch": 1.43, "learning_rate": 3.565389854956327e-05, "loss": 0.0008, "step": 9935 }, { "epoch": 1.43, "learning_rate": 3.5643881721291774e-05, "loss": 0.0013, "step": 9940 }, { "epoch": 1.43, "learning_rate": 3.5633864893020276e-05, "loss": 0.0007, "step": 9945 }, { "epoch": 1.44, "learning_rate": 3.562384806474878e-05, "loss": 0.001, "step": 9950 }, { "epoch": 1.44, "learning_rate": 3.561383123647729e-05, "loss": 0.0011, "step": 9955 }, { "epoch": 1.44, "learning_rate": 3.560381440820579e-05, "loss": 0.0011, "step": 9960 }, { "epoch": 1.44, "learning_rate": 3.559379757993429e-05, "loss": 0.0007, "step": 9965 }, { "epoch": 1.44, "learning_rate": 3.558378075166279e-05, "loss": 0.0007, "step": 9970 }, { "epoch": 1.44, "learning_rate": 3.55737639233913e-05, "loss": 0.0012, "step": 9975 }, { "epoch": 1.44, "learning_rate": 3.55637470951198e-05, "loss": 0.0011, "step": 9980 }, { "epoch": 1.44, "learning_rate": 3.5553730266848304e-05, "loss": 0.0007, "step": 9985 }, { "epoch": 1.44, "learning_rate": 3.5543713438576806e-05, "loss": 0.0012, "step": 9990 }, { "epoch": 1.44, "learning_rate": 3.5533696610305315e-05, "loss": 0.0009, "step": 9995 }, { "epoch": 1.44, "learning_rate": 3.552367978203382e-05, "loss": 0.001, "step": 10000 }, { "epoch": 1.44, "learning_rate": 3.5513662953762325e-05, "loss": 0.001, "step": 10005 }, { "epoch": 1.44, "learning_rate": 3.550364612549083e-05, "loss": 0.0007, "step": 10010 }, { "epoch": 1.44, "learning_rate": 3.5493629297219336e-05, "loss": 0.001, "step": 10015 }, { "epoch": 1.45, "learning_rate": 3.548361246894784e-05, "loss": 0.0009, "step": 10020 }, { "epoch": 1.45, "learning_rate": 3.547359564067634e-05, "loss": 0.0016, "step": 10025 }, { "epoch": 1.45, "learning_rate": 3.546357881240484e-05, "loss": 0.0014, "step": 10030 }, { "epoch": 1.45, "learning_rate": 3.545356198413334e-05, "loss": 0.0011, "step": 10035 }, { "epoch": 1.45, "learning_rate": 3.544354515586185e-05, "loss": 0.0009, "step": 10040 }, { "epoch": 1.45, "learning_rate": 3.5433528327590354e-05, "loss": 0.0012, "step": 10045 }, { "epoch": 1.45, "learning_rate": 3.5423511499318855e-05, "loss": 0.0008, "step": 10050 }, { "epoch": 1.45, "learning_rate": 3.541349467104736e-05, "loss": 0.001, "step": 10055 }, { "epoch": 1.45, "learning_rate": 3.5403477842775866e-05, "loss": 0.001, "step": 10060 }, { "epoch": 1.45, "learning_rate": 3.539346101450437e-05, "loss": 0.001, "step": 10065 }, { "epoch": 1.45, "learning_rate": 3.538344418623287e-05, "loss": 0.0012, "step": 10070 }, { "epoch": 1.45, "learning_rate": 3.537342735796138e-05, "loss": 0.0009, "step": 10075 }, { "epoch": 1.45, "learning_rate": 3.536341052968988e-05, "loss": 0.0006, "step": 10080 }, { "epoch": 1.45, "learning_rate": 3.535339370141839e-05, "loss": 0.0011, "step": 10085 }, { "epoch": 1.46, "learning_rate": 3.534337687314689e-05, "loss": 0.0009, "step": 10090 }, { "epoch": 1.46, "learning_rate": 3.533336004487539e-05, "loss": 0.0009, "step": 10095 }, { "epoch": 1.46, "learning_rate": 3.5323343216603894e-05, "loss": 0.001, "step": 10100 }, { "epoch": 1.46, "learning_rate": 3.53133263883324e-05, "loss": 0.0007, "step": 10105 }, { "epoch": 1.46, "learning_rate": 3.5303309560060905e-05, "loss": 0.001, "step": 10110 }, { "epoch": 1.46, "learning_rate": 3.5293292731789406e-05, "loss": 0.0007, "step": 10115 }, { "epoch": 1.46, "learning_rate": 3.528327590351791e-05, "loss": 0.001, "step": 10120 }, { "epoch": 1.46, "learning_rate": 3.527325907524642e-05, "loss": 0.0011, "step": 10125 }, { "epoch": 1.46, "learning_rate": 3.526324224697492e-05, "loss": 0.0005, "step": 10130 }, { "epoch": 1.46, "learning_rate": 3.525322541870342e-05, "loss": 0.001, "step": 10135 }, { "epoch": 1.46, "learning_rate": 3.524320859043192e-05, "loss": 0.0009, "step": 10140 }, { "epoch": 1.46, "learning_rate": 3.523319176216043e-05, "loss": 0.0009, "step": 10145 }, { "epoch": 1.46, "learning_rate": 3.522317493388893e-05, "loss": 0.0008, "step": 10150 }, { "epoch": 1.46, "learning_rate": 3.521315810561744e-05, "loss": 0.001, "step": 10155 }, { "epoch": 1.47, "learning_rate": 3.520314127734594e-05, "loss": 0.0008, "step": 10160 }, { "epoch": 1.47, "learning_rate": 3.519312444907445e-05, "loss": 0.0015, "step": 10165 }, { "epoch": 1.47, "learning_rate": 3.5183107620802954e-05, "loss": 0.0014, "step": 10170 }, { "epoch": 1.47, "learning_rate": 3.5173090792531456e-05, "loss": 0.0011, "step": 10175 }, { "epoch": 1.47, "learning_rate": 3.516307396425996e-05, "loss": 0.0014, "step": 10180 }, { "epoch": 1.47, "learning_rate": 3.515305713598846e-05, "loss": 0.0011, "step": 10185 }, { "epoch": 1.47, "learning_rate": 3.514304030771697e-05, "loss": 0.0012, "step": 10190 }, { "epoch": 1.47, "learning_rate": 3.513302347944547e-05, "loss": 0.0007, "step": 10195 }, { "epoch": 1.47, "learning_rate": 3.512300665117397e-05, "loss": 0.0011, "step": 10200 }, { "epoch": 1.47, "learning_rate": 3.5112989822902473e-05, "loss": 0.0008, "step": 10205 }, { "epoch": 1.47, "learning_rate": 3.510297299463098e-05, "loss": 0.0012, "step": 10210 }, { "epoch": 1.47, "learning_rate": 3.5092956166359484e-05, "loss": 0.0012, "step": 10215 }, { "epoch": 1.47, "learning_rate": 3.5082939338087986e-05, "loss": 0.0007, "step": 10220 }, { "epoch": 1.47, "learning_rate": 3.5072922509816494e-05, "loss": 0.0011, "step": 10225 }, { "epoch": 1.48, "learning_rate": 3.5062905681545e-05, "loss": 0.0006, "step": 10230 }, { "epoch": 1.48, "learning_rate": 3.5052888853273505e-05, "loss": 0.0007, "step": 10235 }, { "epoch": 1.48, "learning_rate": 3.504287202500201e-05, "loss": 0.0009, "step": 10240 }, { "epoch": 1.48, "learning_rate": 3.503285519673051e-05, "loss": 0.001, "step": 10245 }, { "epoch": 1.48, "learning_rate": 3.502283836845902e-05, "loss": 0.0008, "step": 10250 }, { "epoch": 1.48, "learning_rate": 3.501282154018752e-05, "loss": 0.0009, "step": 10255 }, { "epoch": 1.48, "learning_rate": 3.500280471191602e-05, "loss": 0.001, "step": 10260 }, { "epoch": 1.48, "learning_rate": 3.499278788364452e-05, "loss": 0.0011, "step": 10265 }, { "epoch": 1.48, "learning_rate": 3.4982771055373024e-05, "loss": 0.0009, "step": 10270 }, { "epoch": 1.48, "learning_rate": 3.497275422710153e-05, "loss": 0.0012, "step": 10275 }, { "epoch": 1.48, "learning_rate": 3.4962737398830035e-05, "loss": 0.0007, "step": 10280 }, { "epoch": 1.48, "learning_rate": 3.495272057055854e-05, "loss": 0.0011, "step": 10285 }, { "epoch": 1.48, "learning_rate": 3.494270374228704e-05, "loss": 0.0011, "step": 10290 }, { "epoch": 1.48, "learning_rate": 3.493268691401555e-05, "loss": 0.0011, "step": 10295 }, { "epoch": 1.49, "learning_rate": 3.4922670085744056e-05, "loss": 0.0013, "step": 10300 }, { "epoch": 1.49, "learning_rate": 3.491265325747256e-05, "loss": 0.0015, "step": 10305 }, { "epoch": 1.49, "learning_rate": 3.490263642920106e-05, "loss": 0.0007, "step": 10310 }, { "epoch": 1.49, "learning_rate": 3.489261960092957e-05, "loss": 0.0016, "step": 10315 }, { "epoch": 1.49, "learning_rate": 3.488260277265807e-05, "loss": 0.0006, "step": 10320 }, { "epoch": 1.49, "learning_rate": 3.487258594438657e-05, "loss": 0.0008, "step": 10325 }, { "epoch": 1.49, "learning_rate": 3.4862569116115074e-05, "loss": 0.0013, "step": 10330 }, { "epoch": 1.49, "learning_rate": 3.4852552287843575e-05, "loss": 0.0008, "step": 10335 }, { "epoch": 1.49, "learning_rate": 3.4842535459572084e-05, "loss": 0.0014, "step": 10340 }, { "epoch": 1.49, "learning_rate": 3.4832518631300586e-05, "loss": 0.0013, "step": 10345 }, { "epoch": 1.49, "learning_rate": 3.482250180302909e-05, "loss": 0.0013, "step": 10350 }, { "epoch": 1.49, "learning_rate": 3.481248497475759e-05, "loss": 0.0014, "step": 10355 }, { "epoch": 1.49, "learning_rate": 3.48024681464861e-05, "loss": 0.001, "step": 10360 }, { "epoch": 1.5, "learning_rate": 3.47924513182146e-05, "loss": 0.0007, "step": 10365 }, { "epoch": 1.5, "learning_rate": 3.478243448994311e-05, "loss": 0.0008, "step": 10370 }, { "epoch": 1.5, "learning_rate": 3.477241766167161e-05, "loss": 0.0011, "step": 10375 }, { "epoch": 1.5, "learning_rate": 3.476240083340012e-05, "loss": 0.0013, "step": 10380 }, { "epoch": 1.5, "learning_rate": 3.475238400512862e-05, "loss": 0.0008, "step": 10385 }, { "epoch": 1.5, "learning_rate": 3.474236717685712e-05, "loss": 0.0012, "step": 10390 }, { "epoch": 1.5, "learning_rate": 3.4732350348585625e-05, "loss": 0.0016, "step": 10395 }, { "epoch": 1.5, "learning_rate": 3.472233352031413e-05, "loss": 0.001, "step": 10400 }, { "epoch": 1.5, "learning_rate": 3.4712316692042635e-05, "loss": 0.001, "step": 10405 }, { "epoch": 1.5, "learning_rate": 3.470229986377114e-05, "loss": 0.001, "step": 10410 }, { "epoch": 1.5, "learning_rate": 3.469228303549964e-05, "loss": 0.0009, "step": 10415 }, { "epoch": 1.5, "learning_rate": 3.468226620722814e-05, "loss": 0.001, "step": 10420 }, { "epoch": 1.5, "learning_rate": 3.467224937895665e-05, "loss": 0.001, "step": 10425 }, { "epoch": 1.5, "learning_rate": 3.466223255068515e-05, "loss": 0.0008, "step": 10430 }, { "epoch": 1.51, "learning_rate": 3.465221572241365e-05, "loss": 0.0009, "step": 10435 }, { "epoch": 1.51, "learning_rate": 3.464219889414216e-05, "loss": 0.0007, "step": 10440 }, { "epoch": 1.51, "learning_rate": 3.463218206587066e-05, "loss": 0.0007, "step": 10445 }, { "epoch": 1.51, "learning_rate": 3.462216523759917e-05, "loss": 0.0011, "step": 10450 }, { "epoch": 1.51, "learning_rate": 3.4612148409327674e-05, "loss": 0.0008, "step": 10455 }, { "epoch": 1.51, "learning_rate": 3.4602131581056176e-05, "loss": 0.0011, "step": 10460 }, { "epoch": 1.51, "learning_rate": 3.4592114752784684e-05, "loss": 0.0006, "step": 10465 }, { "epoch": 1.51, "learning_rate": 3.4582097924513186e-05, "loss": 0.001, "step": 10470 }, { "epoch": 1.51, "learning_rate": 3.457208109624169e-05, "loss": 0.0005, "step": 10475 }, { "epoch": 1.51, "learning_rate": 3.456206426797019e-05, "loss": 0.0007, "step": 10480 }, { "epoch": 1.51, "learning_rate": 3.45520474396987e-05, "loss": 0.0009, "step": 10485 }, { "epoch": 1.51, "learning_rate": 3.45420306114272e-05, "loss": 0.0006, "step": 10490 }, { "epoch": 1.51, "learning_rate": 3.45320137831557e-05, "loss": 0.0008, "step": 10495 }, { "epoch": 1.51, "learning_rate": 3.4521996954884204e-05, "loss": 0.001, "step": 10500 }, { "epoch": 1.52, "learning_rate": 3.4511980126612706e-05, "loss": 0.0008, "step": 10505 }, { "epoch": 1.52, "learning_rate": 3.4501963298341214e-05, "loss": 0.0007, "step": 10510 }, { "epoch": 1.52, "learning_rate": 3.4491946470069716e-05, "loss": 0.001, "step": 10515 }, { "epoch": 1.52, "learning_rate": 3.4481929641798225e-05, "loss": 0.0011, "step": 10520 }, { "epoch": 1.52, "learning_rate": 3.447191281352673e-05, "loss": 0.0015, "step": 10525 }, { "epoch": 1.52, "learning_rate": 3.4461895985255235e-05, "loss": 0.0009, "step": 10530 }, { "epoch": 1.52, "learning_rate": 3.445187915698374e-05, "loss": 0.001, "step": 10535 }, { "epoch": 1.52, "learning_rate": 3.444186232871224e-05, "loss": 0.0006, "step": 10540 }, { "epoch": 1.52, "learning_rate": 3.443184550044074e-05, "loss": 0.0009, "step": 10545 }, { "epoch": 1.52, "learning_rate": 3.442182867216925e-05, "loss": 0.0008, "step": 10550 }, { "epoch": 1.52, "learning_rate": 3.441181184389775e-05, "loss": 0.001, "step": 10555 }, { "epoch": 1.52, "learning_rate": 3.440179501562625e-05, "loss": 0.0013, "step": 10560 }, { "epoch": 1.52, "learning_rate": 3.4391778187354755e-05, "loss": 0.0012, "step": 10565 }, { "epoch": 1.52, "learning_rate": 3.438176135908326e-05, "loss": 0.0011, "step": 10570 }, { "epoch": 1.53, "learning_rate": 3.4371744530811765e-05, "loss": 0.0007, "step": 10575 }, { "epoch": 1.53, "learning_rate": 3.436172770254027e-05, "loss": 0.0012, "step": 10580 }, { "epoch": 1.53, "learning_rate": 3.435171087426877e-05, "loss": 0.0008, "step": 10585 }, { "epoch": 1.53, "learning_rate": 3.434169404599728e-05, "loss": 0.0012, "step": 10590 }, { "epoch": 1.53, "learning_rate": 3.433167721772578e-05, "loss": 0.0011, "step": 10595 }, { "epoch": 1.53, "learning_rate": 3.432166038945429e-05, "loss": 0.001, "step": 10600 }, { "epoch": 1.53, "learning_rate": 3.431164356118279e-05, "loss": 0.0009, "step": 10605 }, { "epoch": 1.53, "learning_rate": 3.430162673291129e-05, "loss": 0.0009, "step": 10610 }, { "epoch": 1.53, "learning_rate": 3.42916099046398e-05, "loss": 0.0009, "step": 10615 }, { "epoch": 1.53, "learning_rate": 3.42815930763683e-05, "loss": 0.0009, "step": 10620 }, { "epoch": 1.53, "learning_rate": 3.4271576248096804e-05, "loss": 0.0007, "step": 10625 }, { "epoch": 1.53, "learning_rate": 3.4261559419825306e-05, "loss": 0.001, "step": 10630 }, { "epoch": 1.53, "learning_rate": 3.4251542591553815e-05, "loss": 0.0008, "step": 10635 }, { "epoch": 1.53, "learning_rate": 3.4241525763282316e-05, "loss": 0.0009, "step": 10640 }, { "epoch": 1.54, "learning_rate": 3.423150893501082e-05, "loss": 0.001, "step": 10645 }, { "epoch": 1.54, "learning_rate": 3.422149210673932e-05, "loss": 0.0011, "step": 10650 }, { "epoch": 1.54, "learning_rate": 3.421147527846782e-05, "loss": 0.0013, "step": 10655 }, { "epoch": 1.54, "learning_rate": 3.420145845019633e-05, "loss": 0.0012, "step": 10660 }, { "epoch": 1.54, "learning_rate": 3.419144162192483e-05, "loss": 0.0011, "step": 10665 }, { "epoch": 1.54, "learning_rate": 3.418142479365334e-05, "loss": 0.001, "step": 10670 }, { "epoch": 1.54, "learning_rate": 3.417140796538184e-05, "loss": 0.0008, "step": 10675 }, { "epoch": 1.54, "learning_rate": 3.416139113711035e-05, "loss": 0.0006, "step": 10680 }, { "epoch": 1.54, "learning_rate": 3.415137430883885e-05, "loss": 0.0008, "step": 10685 }, { "epoch": 1.54, "learning_rate": 3.4141357480567355e-05, "loss": 0.0009, "step": 10690 }, { "epoch": 1.54, "learning_rate": 3.413134065229586e-05, "loss": 0.0013, "step": 10695 }, { "epoch": 1.54, "learning_rate": 3.4121323824024366e-05, "loss": 0.0007, "step": 10700 }, { "epoch": 1.54, "learning_rate": 3.411130699575287e-05, "loss": 0.0008, "step": 10705 }, { "epoch": 1.54, "learning_rate": 3.410129016748137e-05, "loss": 0.0008, "step": 10710 }, { "epoch": 1.55, "learning_rate": 3.409127333920987e-05, "loss": 0.001, "step": 10715 }, { "epoch": 1.55, "learning_rate": 3.408125651093838e-05, "loss": 0.0008, "step": 10720 }, { "epoch": 1.55, "learning_rate": 3.407123968266688e-05, "loss": 0.0008, "step": 10725 }, { "epoch": 1.55, "learning_rate": 3.406122285439538e-05, "loss": 0.0011, "step": 10730 }, { "epoch": 1.55, "learning_rate": 3.405120602612389e-05, "loss": 0.0006, "step": 10735 }, { "epoch": 1.55, "learning_rate": 3.4041189197852394e-05, "loss": 0.0008, "step": 10740 }, { "epoch": 1.55, "learning_rate": 3.40311723695809e-05, "loss": 0.0008, "step": 10745 }, { "epoch": 1.55, "learning_rate": 3.4021155541309404e-05, "loss": 0.0012, "step": 10750 }, { "epoch": 1.55, "learning_rate": 3.4011138713037906e-05, "loss": 0.0012, "step": 10755 }, { "epoch": 1.55, "learning_rate": 3.400112188476641e-05, "loss": 0.0011, "step": 10760 }, { "epoch": 1.55, "learning_rate": 3.3991105056494917e-05, "loss": 0.0009, "step": 10765 }, { "epoch": 1.55, "learning_rate": 3.398108822822342e-05, "loss": 0.001, "step": 10770 }, { "epoch": 1.55, "learning_rate": 3.397107139995192e-05, "loss": 0.0013, "step": 10775 }, { "epoch": 1.55, "learning_rate": 3.396105457168042e-05, "loss": 0.0011, "step": 10780 }, { "epoch": 1.56, "learning_rate": 3.395103774340893e-05, "loss": 0.0008, "step": 10785 }, { "epoch": 1.56, "learning_rate": 3.394102091513743e-05, "loss": 0.0007, "step": 10790 }, { "epoch": 1.56, "learning_rate": 3.3931004086865934e-05, "loss": 0.0005, "step": 10795 }, { "epoch": 1.56, "learning_rate": 3.3920987258594436e-05, "loss": 0.001, "step": 10800 }, { "epoch": 1.56, "learning_rate": 3.3910970430322945e-05, "loss": 0.0012, "step": 10805 }, { "epoch": 1.56, "learning_rate": 3.390095360205145e-05, "loss": 0.0009, "step": 10810 }, { "epoch": 1.56, "learning_rate": 3.3890936773779955e-05, "loss": 0.0009, "step": 10815 }, { "epoch": 1.56, "learning_rate": 3.388091994550846e-05, "loss": 0.001, "step": 10820 }, { "epoch": 1.56, "learning_rate": 3.387090311723696e-05, "loss": 0.0014, "step": 10825 }, { "epoch": 1.56, "learning_rate": 3.386088628896547e-05, "loss": 0.0007, "step": 10830 }, { "epoch": 1.56, "learning_rate": 3.385086946069397e-05, "loss": 0.0007, "step": 10835 }, { "epoch": 1.56, "learning_rate": 3.384085263242247e-05, "loss": 0.0012, "step": 10840 }, { "epoch": 1.56, "learning_rate": 3.383083580415097e-05, "loss": 0.0008, "step": 10845 }, { "epoch": 1.56, "learning_rate": 3.382081897587948e-05, "loss": 0.0008, "step": 10850 }, { "epoch": 1.57, "learning_rate": 3.3810802147607983e-05, "loss": 0.0011, "step": 10855 }, { "epoch": 1.57, "learning_rate": 3.3800785319336485e-05, "loss": 0.0012, "step": 10860 }, { "epoch": 1.57, "learning_rate": 3.379076849106499e-05, "loss": 0.0008, "step": 10865 }, { "epoch": 1.57, "learning_rate": 3.3780751662793496e-05, "loss": 0.0013, "step": 10870 }, { "epoch": 1.57, "learning_rate": 3.3770734834522e-05, "loss": 0.0009, "step": 10875 }, { "epoch": 1.57, "learning_rate": 3.37607180062505e-05, "loss": 0.0009, "step": 10880 }, { "epoch": 1.57, "learning_rate": 3.375070117797901e-05, "loss": 0.001, "step": 10885 }, { "epoch": 1.57, "learning_rate": 3.374068434970751e-05, "loss": 0.001, "step": 10890 }, { "epoch": 1.57, "learning_rate": 3.373066752143602e-05, "loss": 0.0008, "step": 10895 }, { "epoch": 1.57, "learning_rate": 3.372065069316452e-05, "loss": 0.0007, "step": 10900 }, { "epoch": 1.57, "learning_rate": 3.371063386489302e-05, "loss": 0.0009, "step": 10905 }, { "epoch": 1.57, "learning_rate": 3.3700617036621524e-05, "loss": 0.0007, "step": 10910 }, { "epoch": 1.57, "learning_rate": 3.369060020835003e-05, "loss": 0.0007, "step": 10915 }, { "epoch": 1.58, "learning_rate": 3.3680583380078535e-05, "loss": 0.0009, "step": 10920 }, { "epoch": 1.58, "learning_rate": 3.3670566551807036e-05, "loss": 0.0008, "step": 10925 }, { "epoch": 1.58, "learning_rate": 3.366054972353554e-05, "loss": 0.0006, "step": 10930 }, { "epoch": 1.58, "learning_rate": 3.365053289526405e-05, "loss": 0.0011, "step": 10935 }, { "epoch": 1.58, "learning_rate": 3.364051606699255e-05, "loss": 0.0009, "step": 10940 }, { "epoch": 1.58, "learning_rate": 3.363049923872105e-05, "loss": 0.0009, "step": 10945 }, { "epoch": 1.58, "learning_rate": 3.362048241044955e-05, "loss": 0.0008, "step": 10950 }, { "epoch": 1.58, "learning_rate": 3.361046558217806e-05, "loss": 0.0013, "step": 10955 }, { "epoch": 1.58, "learning_rate": 3.360044875390656e-05, "loss": 0.001, "step": 10960 }, { "epoch": 1.58, "learning_rate": 3.359043192563507e-05, "loss": 0.0006, "step": 10965 }, { "epoch": 1.58, "learning_rate": 3.358041509736357e-05, "loss": 0.0008, "step": 10970 }, { "epoch": 1.58, "learning_rate": 3.3570398269092075e-05, "loss": 0.0009, "step": 10975 }, { "epoch": 1.58, "learning_rate": 3.3560381440820584e-05, "loss": 0.0008, "step": 10980 }, { "epoch": 1.58, "learning_rate": 3.3550364612549086e-05, "loss": 0.0014, "step": 10985 }, { "epoch": 1.59, "learning_rate": 3.354034778427759e-05, "loss": 0.0009, "step": 10990 }, { "epoch": 1.59, "learning_rate": 3.353033095600609e-05, "loss": 0.0008, "step": 10995 }, { "epoch": 1.59, "learning_rate": 3.35203141277346e-05, "loss": 0.0011, "step": 11000 }, { "epoch": 1.59, "learning_rate": 3.35102972994631e-05, "loss": 0.001, "step": 11005 }, { "epoch": 1.59, "learning_rate": 3.35002804711916e-05, "loss": 0.001, "step": 11010 }, { "epoch": 1.59, "learning_rate": 3.34902636429201e-05, "loss": 0.0009, "step": 11015 }, { "epoch": 1.59, "learning_rate": 3.348024681464861e-05, "loss": 0.0008, "step": 11020 }, { "epoch": 1.59, "learning_rate": 3.3470229986377114e-05, "loss": 0.0011, "step": 11025 }, { "epoch": 1.59, "learning_rate": 3.3460213158105616e-05, "loss": 0.0008, "step": 11030 }, { "epoch": 1.59, "learning_rate": 3.3450196329834124e-05, "loss": 0.0006, "step": 11035 }, { "epoch": 1.59, "learning_rate": 3.344017950156263e-05, "loss": 0.0012, "step": 11040 }, { "epoch": 1.59, "learning_rate": 3.3430162673291135e-05, "loss": 0.0012, "step": 11045 }, { "epoch": 1.59, "learning_rate": 3.3420145845019637e-05, "loss": 0.0012, "step": 11050 }, { "epoch": 1.59, "learning_rate": 3.341012901674814e-05, "loss": 0.001, "step": 11055 }, { "epoch": 1.6, "learning_rate": 3.340011218847664e-05, "loss": 0.0011, "step": 11060 }, { "epoch": 1.6, "learning_rate": 3.339009536020515e-05, "loss": 0.0011, "step": 11065 }, { "epoch": 1.6, "learning_rate": 3.338007853193365e-05, "loss": 0.001, "step": 11070 }, { "epoch": 1.6, "learning_rate": 3.337006170366215e-05, "loss": 0.0006, "step": 11075 }, { "epoch": 1.6, "learning_rate": 3.3360044875390654e-05, "loss": 0.0011, "step": 11080 }, { "epoch": 1.6, "learning_rate": 3.335002804711916e-05, "loss": 0.0008, "step": 11085 }, { "epoch": 1.6, "learning_rate": 3.3340011218847665e-05, "loss": 0.0014, "step": 11090 }, { "epoch": 1.6, "learning_rate": 3.332999439057617e-05, "loss": 0.0012, "step": 11095 }, { "epoch": 1.6, "learning_rate": 3.331997756230467e-05, "loss": 0.0006, "step": 11100 }, { "epoch": 1.6, "learning_rate": 3.330996073403318e-05, "loss": 0.0007, "step": 11105 }, { "epoch": 1.6, "learning_rate": 3.3299943905761686e-05, "loss": 0.0013, "step": 11110 }, { "epoch": 1.6, "learning_rate": 3.328992707749019e-05, "loss": 0.0015, "step": 11115 }, { "epoch": 1.6, "learning_rate": 3.327991024921869e-05, "loss": 0.0013, "step": 11120 }, { "epoch": 1.6, "learning_rate": 3.326989342094719e-05, "loss": 0.0012, "step": 11125 }, { "epoch": 1.61, "learning_rate": 3.32598765926757e-05, "loss": 0.0008, "step": 11130 }, { "epoch": 1.61, "learning_rate": 3.32498597644042e-05, "loss": 0.0011, "step": 11135 }, { "epoch": 1.61, "learning_rate": 3.3239842936132703e-05, "loss": 0.0008, "step": 11140 }, { "epoch": 1.61, "learning_rate": 3.3229826107861205e-05, "loss": 0.0006, "step": 11145 }, { "epoch": 1.61, "learning_rate": 3.3219809279589714e-05, "loss": 0.0009, "step": 11150 }, { "epoch": 1.61, "learning_rate": 3.3209792451318216e-05, "loss": 0.0009, "step": 11155 }, { "epoch": 1.61, "learning_rate": 3.319977562304672e-05, "loss": 0.001, "step": 11160 }, { "epoch": 1.61, "learning_rate": 3.318975879477522e-05, "loss": 0.0013, "step": 11165 }, { "epoch": 1.61, "learning_rate": 3.317974196650373e-05, "loss": 0.0009, "step": 11170 }, { "epoch": 1.61, "learning_rate": 3.316972513823223e-05, "loss": 0.0011, "step": 11175 }, { "epoch": 1.61, "learning_rate": 3.315970830996074e-05, "loss": 0.0011, "step": 11180 }, { "epoch": 1.61, "learning_rate": 3.314969148168924e-05, "loss": 0.0009, "step": 11185 }, { "epoch": 1.61, "learning_rate": 3.313967465341775e-05, "loss": 0.0012, "step": 11190 }, { "epoch": 1.61, "learning_rate": 3.312965782514625e-05, "loss": 0.0015, "step": 11195 }, { "epoch": 1.62, "learning_rate": 3.311964099687475e-05, "loss": 0.001, "step": 11200 }, { "epoch": 1.62, "learning_rate": 3.3109624168603255e-05, "loss": 0.0009, "step": 11205 }, { "epoch": 1.62, "learning_rate": 3.3099607340331756e-05, "loss": 0.0013, "step": 11210 }, { "epoch": 1.62, "learning_rate": 3.3089590512060265e-05, "loss": 0.0011, "step": 11215 }, { "epoch": 1.62, "learning_rate": 3.307957368378877e-05, "loss": 0.0009, "step": 11220 }, { "epoch": 1.62, "learning_rate": 3.306955685551727e-05, "loss": 0.001, "step": 11225 }, { "epoch": 1.62, "learning_rate": 3.305954002724577e-05, "loss": 0.0007, "step": 11230 }, { "epoch": 1.62, "learning_rate": 3.304952319897428e-05, "loss": 0.0007, "step": 11235 }, { "epoch": 1.62, "learning_rate": 3.303950637070278e-05, "loss": 0.0009, "step": 11240 }, { "epoch": 1.62, "learning_rate": 3.302948954243128e-05, "loss": 0.0008, "step": 11245 }, { "epoch": 1.62, "learning_rate": 3.301947271415979e-05, "loss": 0.0007, "step": 11250 }, { "epoch": 1.62, "learning_rate": 3.300945588588829e-05, "loss": 0.0011, "step": 11255 }, { "epoch": 1.62, "learning_rate": 3.29994390576168e-05, "loss": 0.0008, "step": 11260 }, { "epoch": 1.62, "learning_rate": 3.2989422229345304e-05, "loss": 0.001, "step": 11265 }, { "epoch": 1.63, "learning_rate": 3.2979405401073806e-05, "loss": 0.0011, "step": 11270 }, { "epoch": 1.63, "learning_rate": 3.2969388572802314e-05, "loss": 0.0012, "step": 11275 }, { "epoch": 1.63, "learning_rate": 3.2959371744530816e-05, "loss": 0.0017, "step": 11280 }, { "epoch": 1.63, "learning_rate": 3.294935491625932e-05, "loss": 0.001, "step": 11285 }, { "epoch": 1.63, "learning_rate": 3.293933808798782e-05, "loss": 0.001, "step": 11290 }, { "epoch": 1.63, "learning_rate": 3.292932125971632e-05, "loss": 0.0017, "step": 11295 }, { "epoch": 1.63, "learning_rate": 3.291930443144483e-05, "loss": 0.0013, "step": 11300 }, { "epoch": 1.63, "learning_rate": 3.290928760317333e-05, "loss": 0.001, "step": 11305 }, { "epoch": 1.63, "learning_rate": 3.2899270774901834e-05, "loss": 0.0008, "step": 11310 }, { "epoch": 1.63, "learning_rate": 3.2889253946630336e-05, "loss": 0.0005, "step": 11315 }, { "epoch": 1.63, "learning_rate": 3.2879237118358844e-05, "loss": 0.0012, "step": 11320 }, { "epoch": 1.63, "learning_rate": 3.2869220290087346e-05, "loss": 0.001, "step": 11325 }, { "epoch": 1.63, "learning_rate": 3.2859203461815855e-05, "loss": 0.0009, "step": 11330 }, { "epoch": 1.63, "learning_rate": 3.2849186633544357e-05, "loss": 0.0011, "step": 11335 }, { "epoch": 1.64, "learning_rate": 3.2839169805272865e-05, "loss": 0.0008, "step": 11340 }, { "epoch": 1.64, "learning_rate": 3.282915297700137e-05, "loss": 0.001, "step": 11345 }, { "epoch": 1.64, "learning_rate": 3.281913614872987e-05, "loss": 0.0008, "step": 11350 }, { "epoch": 1.64, "learning_rate": 3.280911932045837e-05, "loss": 0.0012, "step": 11355 }, { "epoch": 1.64, "learning_rate": 3.279910249218687e-05, "loss": 0.001, "step": 11360 }, { "epoch": 1.64, "learning_rate": 3.278908566391538e-05, "loss": 0.0012, "step": 11365 }, { "epoch": 1.64, "learning_rate": 3.277906883564388e-05, "loss": 0.0008, "step": 11370 }, { "epoch": 1.64, "learning_rate": 3.2769052007372385e-05, "loss": 0.0014, "step": 11375 }, { "epoch": 1.64, "learning_rate": 3.275903517910089e-05, "loss": 0.0008, "step": 11380 }, { "epoch": 1.64, "learning_rate": 3.2749018350829395e-05, "loss": 0.001, "step": 11385 }, { "epoch": 1.64, "learning_rate": 3.27390015225579e-05, "loss": 0.0012, "step": 11390 }, { "epoch": 1.64, "learning_rate": 3.27289846942864e-05, "loss": 0.0009, "step": 11395 }, { "epoch": 1.64, "learning_rate": 3.271896786601491e-05, "loss": 0.0008, "step": 11400 }, { "epoch": 1.65, "learning_rate": 3.270895103774341e-05, "loss": 0.0008, "step": 11405 }, { "epoch": 1.65, "learning_rate": 3.269893420947192e-05, "loss": 0.001, "step": 11410 }, { "epoch": 1.65, "learning_rate": 3.268891738120042e-05, "loss": 0.0012, "step": 11415 }, { "epoch": 1.65, "learning_rate": 3.267890055292892e-05, "loss": 0.001, "step": 11420 }, { "epoch": 1.65, "learning_rate": 3.266888372465743e-05, "loss": 0.0008, "step": 11425 }, { "epoch": 1.65, "learning_rate": 3.265886689638593e-05, "loss": 0.0009, "step": 11430 }, { "epoch": 1.65, "learning_rate": 3.2648850068114434e-05, "loss": 0.001, "step": 11435 }, { "epoch": 1.65, "learning_rate": 3.2638833239842936e-05, "loss": 0.0011, "step": 11440 }, { "epoch": 1.65, "learning_rate": 3.262881641157144e-05, "loss": 0.0014, "step": 11445 }, { "epoch": 1.65, "learning_rate": 3.2618799583299946e-05, "loss": 0.0012, "step": 11450 }, { "epoch": 1.65, "learning_rate": 3.260878275502845e-05, "loss": 0.0011, "step": 11455 }, { "epoch": 1.65, "learning_rate": 3.259876592675695e-05, "loss": 0.0009, "step": 11460 }, { "epoch": 1.65, "learning_rate": 3.258874909848545e-05, "loss": 0.0007, "step": 11465 }, { "epoch": 1.65, "learning_rate": 3.257873227021396e-05, "loss": 0.0008, "step": 11470 }, { "epoch": 1.66, "learning_rate": 3.256871544194247e-05, "loss": 0.0009, "step": 11475 }, { "epoch": 1.66, "learning_rate": 3.255869861367097e-05, "loss": 0.001, "step": 11480 }, { "epoch": 1.66, "learning_rate": 3.254868178539947e-05, "loss": 0.0007, "step": 11485 }, { "epoch": 1.66, "learning_rate": 3.253866495712798e-05, "loss": 0.0006, "step": 11490 }, { "epoch": 1.66, "learning_rate": 3.252864812885648e-05, "loss": 0.0008, "step": 11495 }, { "epoch": 1.66, "learning_rate": 3.2518631300584985e-05, "loss": 0.0009, "step": 11500 }, { "epoch": 1.66, "learning_rate": 3.250861447231349e-05, "loss": 0.001, "step": 11505 }, { "epoch": 1.66, "learning_rate": 3.2498597644041995e-05, "loss": 0.001, "step": 11510 }, { "epoch": 1.66, "learning_rate": 3.24885808157705e-05, "loss": 0.0007, "step": 11515 }, { "epoch": 1.66, "learning_rate": 3.2478563987499e-05, "loss": 0.0011, "step": 11520 }, { "epoch": 1.66, "learning_rate": 3.24685471592275e-05, "loss": 0.0013, "step": 11525 }, { "epoch": 1.66, "learning_rate": 3.2458530330956e-05, "loss": 0.0009, "step": 11530 }, { "epoch": 1.66, "learning_rate": 3.244851350268451e-05, "loss": 0.001, "step": 11535 }, { "epoch": 1.66, "learning_rate": 3.243849667441301e-05, "loss": 0.0007, "step": 11540 }, { "epoch": 1.67, "learning_rate": 3.242847984614152e-05, "loss": 0.0015, "step": 11545 }, { "epoch": 1.67, "learning_rate": 3.2418463017870024e-05, "loss": 0.0008, "step": 11550 }, { "epoch": 1.67, "learning_rate": 3.240844618959853e-05, "loss": 0.0011, "step": 11555 }, { "epoch": 1.67, "learning_rate": 3.2398429361327034e-05, "loss": 0.0011, "step": 11560 }, { "epoch": 1.67, "learning_rate": 3.2388412533055536e-05, "loss": 0.0007, "step": 11565 }, { "epoch": 1.67, "learning_rate": 3.237839570478404e-05, "loss": 0.0013, "step": 11570 }, { "epoch": 1.67, "learning_rate": 3.2368378876512546e-05, "loss": 0.0017, "step": 11575 }, { "epoch": 1.67, "learning_rate": 3.235836204824105e-05, "loss": 0.0008, "step": 11580 }, { "epoch": 1.67, "learning_rate": 3.234834521996955e-05, "loss": 0.001, "step": 11585 }, { "epoch": 1.67, "learning_rate": 3.233832839169805e-05, "loss": 0.0008, "step": 11590 }, { "epoch": 1.67, "learning_rate": 3.2328311563426554e-05, "loss": 0.001, "step": 11595 }, { "epoch": 1.67, "learning_rate": 3.231829473515506e-05, "loss": 0.0006, "step": 11600 }, { "epoch": 1.67, "learning_rate": 3.2308277906883564e-05, "loss": 0.001, "step": 11605 }, { "epoch": 1.67, "learning_rate": 3.2298261078612066e-05, "loss": 0.0009, "step": 11610 }, { "epoch": 1.68, "learning_rate": 3.2288244250340575e-05, "loss": 0.0007, "step": 11615 }, { "epoch": 1.68, "learning_rate": 3.2278227422069077e-05, "loss": 0.0015, "step": 11620 }, { "epoch": 1.68, "learning_rate": 3.2268210593797585e-05, "loss": 0.001, "step": 11625 }, { "epoch": 1.68, "learning_rate": 3.225819376552609e-05, "loss": 0.001, "step": 11630 }, { "epoch": 1.68, "learning_rate": 3.224817693725459e-05, "loss": 0.0012, "step": 11635 }, { "epoch": 1.68, "learning_rate": 3.22381601089831e-05, "loss": 0.0009, "step": 11640 }, { "epoch": 1.68, "learning_rate": 3.22281432807116e-05, "loss": 0.0007, "step": 11645 }, { "epoch": 1.68, "learning_rate": 3.22181264524401e-05, "loss": 0.0008, "step": 11650 }, { "epoch": 1.68, "learning_rate": 3.22081096241686e-05, "loss": 0.0011, "step": 11655 }, { "epoch": 1.68, "learning_rate": 3.219809279589711e-05, "loss": 0.001, "step": 11660 }, { "epoch": 1.68, "learning_rate": 3.2188075967625613e-05, "loss": 0.0011, "step": 11665 }, { "epoch": 1.68, "learning_rate": 3.2178059139354115e-05, "loss": 0.0016, "step": 11670 }, { "epoch": 1.68, "learning_rate": 3.216804231108262e-05, "loss": 0.0007, "step": 11675 }, { "epoch": 1.68, "learning_rate": 3.215802548281112e-05, "loss": 0.0012, "step": 11680 }, { "epoch": 1.69, "learning_rate": 3.214800865453963e-05, "loss": 0.0008, "step": 11685 }, { "epoch": 1.69, "learning_rate": 3.213799182626813e-05, "loss": 0.0009, "step": 11690 }, { "epoch": 1.69, "learning_rate": 3.212797499799664e-05, "loss": 0.001, "step": 11695 }, { "epoch": 1.69, "learning_rate": 3.211795816972514e-05, "loss": 0.0009, "step": 11700 }, { "epoch": 1.69, "learning_rate": 3.210794134145365e-05, "loss": 0.0008, "step": 11705 }, { "epoch": 1.69, "learning_rate": 3.209792451318215e-05, "loss": 0.0007, "step": 11710 }, { "epoch": 1.69, "learning_rate": 3.208790768491065e-05, "loss": 0.0007, "step": 11715 }, { "epoch": 1.69, "learning_rate": 3.2077890856639154e-05, "loss": 0.0008, "step": 11720 }, { "epoch": 1.69, "learning_rate": 3.206787402836766e-05, "loss": 0.0013, "step": 11725 }, { "epoch": 1.69, "learning_rate": 3.2057857200096164e-05, "loss": 0.001, "step": 11730 }, { "epoch": 1.69, "learning_rate": 3.2047840371824666e-05, "loss": 0.0007, "step": 11735 }, { "epoch": 1.69, "learning_rate": 3.203782354355317e-05, "loss": 0.001, "step": 11740 }, { "epoch": 1.69, "learning_rate": 3.202780671528168e-05, "loss": 0.0007, "step": 11745 }, { "epoch": 1.69, "learning_rate": 3.201778988701018e-05, "loss": 0.0006, "step": 11750 }, { "epoch": 1.7, "learning_rate": 3.200777305873868e-05, "loss": 0.001, "step": 11755 }, { "epoch": 1.7, "learning_rate": 3.199775623046718e-05, "loss": 0.0011, "step": 11760 }, { "epoch": 1.7, "learning_rate": 3.198773940219569e-05, "loss": 0.0006, "step": 11765 }, { "epoch": 1.7, "learning_rate": 3.197772257392419e-05, "loss": 0.001, "step": 11770 }, { "epoch": 1.7, "learning_rate": 3.19677057456527e-05, "loss": 0.0014, "step": 11775 }, { "epoch": 1.7, "learning_rate": 3.19576889173812e-05, "loss": 0.0008, "step": 11780 }, { "epoch": 1.7, "learning_rate": 3.1947672089109705e-05, "loss": 0.0011, "step": 11785 }, { "epoch": 1.7, "learning_rate": 3.1937655260838214e-05, "loss": 0.0011, "step": 11790 }, { "epoch": 1.7, "learning_rate": 3.1927638432566715e-05, "loss": 0.0011, "step": 11795 }, { "epoch": 1.7, "learning_rate": 3.191762160429522e-05, "loss": 0.0011, "step": 11800 }, { "epoch": 1.7, "learning_rate": 3.190760477602372e-05, "loss": 0.0016, "step": 11805 }, { "epoch": 1.7, "learning_rate": 3.189758794775223e-05, "loss": 0.0008, "step": 11810 }, { "epoch": 1.7, "learning_rate": 3.188757111948073e-05, "loss": 0.0009, "step": 11815 }, { "epoch": 1.7, "learning_rate": 3.187755429120923e-05, "loss": 0.001, "step": 11820 }, { "epoch": 1.71, "learning_rate": 3.186753746293773e-05, "loss": 0.0007, "step": 11825 }, { "epoch": 1.71, "learning_rate": 3.1857520634666235e-05, "loss": 0.0008, "step": 11830 }, { "epoch": 1.71, "learning_rate": 3.1847503806394744e-05, "loss": 0.0006, "step": 11835 }, { "epoch": 1.71, "learning_rate": 3.1837486978123246e-05, "loss": 0.0006, "step": 11840 }, { "epoch": 1.71, "learning_rate": 3.1827470149851754e-05, "loss": 0.0012, "step": 11845 }, { "epoch": 1.71, "learning_rate": 3.1817453321580256e-05, "loss": 0.0009, "step": 11850 }, { "epoch": 1.71, "learning_rate": 3.1807436493308765e-05, "loss": 0.0007, "step": 11855 }, { "epoch": 1.71, "learning_rate": 3.1797419665037266e-05, "loss": 0.0011, "step": 11860 }, { "epoch": 1.71, "learning_rate": 3.178740283676577e-05, "loss": 0.0009, "step": 11865 }, { "epoch": 1.71, "learning_rate": 3.177738600849427e-05, "loss": 0.0007, "step": 11870 }, { "epoch": 1.71, "learning_rate": 3.176736918022278e-05, "loss": 0.0009, "step": 11875 }, { "epoch": 1.71, "learning_rate": 3.175735235195128e-05, "loss": 0.0011, "step": 11880 }, { "epoch": 1.71, "learning_rate": 3.174733552367978e-05, "loss": 0.0006, "step": 11885 }, { "epoch": 1.71, "learning_rate": 3.1737318695408284e-05, "loss": 0.0008, "step": 11890 }, { "epoch": 1.72, "learning_rate": 3.172730186713679e-05, "loss": 0.0013, "step": 11895 }, { "epoch": 1.72, "learning_rate": 3.1717285038865295e-05, "loss": 0.0009, "step": 11900 }, { "epoch": 1.72, "learning_rate": 3.1707268210593797e-05, "loss": 0.0009, "step": 11905 }, { "epoch": 1.72, "learning_rate": 3.16972513823223e-05, "loss": 0.0008, "step": 11910 }, { "epoch": 1.72, "learning_rate": 3.168723455405081e-05, "loss": 0.0007, "step": 11915 }, { "epoch": 1.72, "learning_rate": 3.1677217725779316e-05, "loss": 0.0008, "step": 11920 }, { "epoch": 1.72, "learning_rate": 3.166720089750782e-05, "loss": 0.001, "step": 11925 }, { "epoch": 1.72, "learning_rate": 3.165718406923632e-05, "loss": 0.001, "step": 11930 }, { "epoch": 1.72, "learning_rate": 3.164716724096482e-05, "loss": 0.001, "step": 11935 }, { "epoch": 1.72, "learning_rate": 3.163715041269333e-05, "loss": 0.0009, "step": 11940 }, { "epoch": 1.72, "learning_rate": 3.162713358442183e-05, "loss": 0.0008, "step": 11945 }, { "epoch": 1.72, "learning_rate": 3.161711675615033e-05, "loss": 0.0007, "step": 11950 }, { "epoch": 1.72, "learning_rate": 3.1607099927878835e-05, "loss": 0.0007, "step": 11955 }, { "epoch": 1.73, "learning_rate": 3.1597083099607344e-05, "loss": 0.001, "step": 11960 }, { "epoch": 1.73, "learning_rate": 3.1587066271335846e-05, "loss": 0.0009, "step": 11965 }, { "epoch": 1.73, "learning_rate": 3.157704944306435e-05, "loss": 0.0009, "step": 11970 }, { "epoch": 1.73, "learning_rate": 3.156703261479285e-05, "loss": 0.0011, "step": 11975 }, { "epoch": 1.73, "learning_rate": 3.155701578652136e-05, "loss": 0.0009, "step": 11980 }, { "epoch": 1.73, "learning_rate": 3.154699895824986e-05, "loss": 0.0011, "step": 11985 }, { "epoch": 1.73, "learning_rate": 3.153698212997837e-05, "loss": 0.0008, "step": 11990 }, { "epoch": 1.73, "learning_rate": 3.152696530170687e-05, "loss": 0.0009, "step": 11995 }, { "epoch": 1.73, "learning_rate": 3.151694847343537e-05, "loss": 0.001, "step": 12000 }, { "epoch": 1.73, "learning_rate": 3.150693164516388e-05, "loss": 0.0011, "step": 12005 }, { "epoch": 1.73, "learning_rate": 3.149691481689238e-05, "loss": 0.0011, "step": 12010 }, { "epoch": 1.73, "learning_rate": 3.1486897988620884e-05, "loss": 0.0008, "step": 12015 }, { "epoch": 1.73, "learning_rate": 3.1476881160349386e-05, "loss": 0.0008, "step": 12020 }, { "epoch": 1.73, "learning_rate": 3.1466864332077895e-05, "loss": 0.0009, "step": 12025 }, { "epoch": 1.74, "learning_rate": 3.14568475038064e-05, "loss": 0.0011, "step": 12030 }, { "epoch": 1.74, "learning_rate": 3.14468306755349e-05, "loss": 0.001, "step": 12035 }, { "epoch": 1.74, "learning_rate": 3.14368138472634e-05, "loss": 0.001, "step": 12040 }, { "epoch": 1.74, "learning_rate": 3.142679701899191e-05, "loss": 0.0012, "step": 12045 }, { "epoch": 1.74, "learning_rate": 3.141678019072041e-05, "loss": 0.0009, "step": 12050 }, { "epoch": 1.74, "learning_rate": 3.140676336244891e-05, "loss": 0.0006, "step": 12055 }, { "epoch": 1.74, "learning_rate": 3.139674653417742e-05, "loss": 0.0009, "step": 12060 }, { "epoch": 1.74, "learning_rate": 3.138672970590592e-05, "loss": 0.0007, "step": 12065 }, { "epoch": 1.74, "learning_rate": 3.137671287763443e-05, "loss": 0.001, "step": 12070 }, { "epoch": 1.74, "learning_rate": 3.1366696049362934e-05, "loss": 0.0009, "step": 12075 }, { "epoch": 1.74, "learning_rate": 3.1356679221091435e-05, "loss": 0.0011, "step": 12080 }, { "epoch": 1.74, "learning_rate": 3.134666239281994e-05, "loss": 0.0007, "step": 12085 }, { "epoch": 1.74, "learning_rate": 3.1336645564548446e-05, "loss": 0.0011, "step": 12090 }, { "epoch": 1.74, "learning_rate": 3.132662873627695e-05, "loss": 0.0009, "step": 12095 }, { "epoch": 1.75, "learning_rate": 3.131661190800545e-05, "loss": 0.0009, "step": 12100 }, { "epoch": 1.75, "learning_rate": 3.130659507973395e-05, "loss": 0.0006, "step": 12105 }, { "epoch": 1.75, "learning_rate": 3.129657825146246e-05, "loss": 0.0007, "step": 12110 }, { "epoch": 1.75, "learning_rate": 3.128656142319096e-05, "loss": 0.0007, "step": 12115 }, { "epoch": 1.75, "learning_rate": 3.1276544594919464e-05, "loss": 0.001, "step": 12120 }, { "epoch": 1.75, "learning_rate": 3.1266527766647966e-05, "loss": 0.0008, "step": 12125 }, { "epoch": 1.75, "learning_rate": 3.1256510938376474e-05, "loss": 0.001, "step": 12130 }, { "epoch": 1.75, "learning_rate": 3.1246494110104976e-05, "loss": 0.0012, "step": 12135 }, { "epoch": 1.75, "learning_rate": 3.1236477281833485e-05, "loss": 0.001, "step": 12140 }, { "epoch": 1.75, "learning_rate": 3.1226460453561986e-05, "loss": 0.0009, "step": 12145 }, { "epoch": 1.75, "learning_rate": 3.121644362529049e-05, "loss": 0.001, "step": 12150 }, { "epoch": 1.75, "learning_rate": 3.1206426797019e-05, "loss": 0.0011, "step": 12155 }, { "epoch": 1.75, "learning_rate": 3.11964099687475e-05, "loss": 0.0007, "step": 12160 }, { "epoch": 1.75, "learning_rate": 3.1186393140476e-05, "loss": 0.0011, "step": 12165 }, { "epoch": 1.76, "learning_rate": 3.11763763122045e-05, "loss": 0.0006, "step": 12170 }, { "epoch": 1.76, "learning_rate": 3.116635948393301e-05, "loss": 0.001, "step": 12175 }, { "epoch": 1.76, "learning_rate": 3.115634265566151e-05, "loss": 0.0007, "step": 12180 }, { "epoch": 1.76, "learning_rate": 3.1146325827390015e-05, "loss": 0.0007, "step": 12185 }, { "epoch": 1.76, "learning_rate": 3.1136308999118517e-05, "loss": 0.0013, "step": 12190 }, { "epoch": 1.76, "learning_rate": 3.1126292170847025e-05, "loss": 0.0009, "step": 12195 }, { "epoch": 1.76, "learning_rate": 3.111627534257553e-05, "loss": 0.0008, "step": 12200 }, { "epoch": 1.76, "learning_rate": 3.110625851430403e-05, "loss": 0.0011, "step": 12205 }, { "epoch": 1.76, "learning_rate": 3.109624168603254e-05, "loss": 0.0012, "step": 12210 }, { "epoch": 1.76, "learning_rate": 3.1086224857761046e-05, "loss": 0.0008, "step": 12215 }, { "epoch": 1.76, "learning_rate": 3.107620802948955e-05, "loss": 0.001, "step": 12220 }, { "epoch": 1.76, "learning_rate": 3.106619120121805e-05, "loss": 0.0008, "step": 12225 }, { "epoch": 1.76, "learning_rate": 3.105617437294655e-05, "loss": 0.0009, "step": 12230 }, { "epoch": 1.76, "learning_rate": 3.104615754467505e-05, "loss": 0.0009, "step": 12235 }, { "epoch": 1.77, "learning_rate": 3.103614071640356e-05, "loss": 0.0008, "step": 12240 }, { "epoch": 1.77, "learning_rate": 3.1026123888132064e-05, "loss": 0.001, "step": 12245 }, { "epoch": 1.77, "learning_rate": 3.1016107059860566e-05, "loss": 0.001, "step": 12250 }, { "epoch": 1.77, "learning_rate": 3.100609023158907e-05, "loss": 0.001, "step": 12255 }, { "epoch": 1.77, "learning_rate": 3.0996073403317576e-05, "loss": 0.0009, "step": 12260 }, { "epoch": 1.77, "learning_rate": 3.098605657504608e-05, "loss": 0.0007, "step": 12265 }, { "epoch": 1.77, "learning_rate": 3.097603974677458e-05, "loss": 0.0009, "step": 12270 }, { "epoch": 1.77, "learning_rate": 3.096602291850308e-05, "loss": 0.0011, "step": 12275 }, { "epoch": 1.77, "learning_rate": 3.095600609023159e-05, "loss": 0.0009, "step": 12280 }, { "epoch": 1.77, "learning_rate": 3.09459892619601e-05, "loss": 0.001, "step": 12285 }, { "epoch": 1.77, "learning_rate": 3.09359724336886e-05, "loss": 0.0011, "step": 12290 }, { "epoch": 1.77, "learning_rate": 3.09259556054171e-05, "loss": 0.0009, "step": 12295 }, { "epoch": 1.77, "learning_rate": 3.091593877714561e-05, "loss": 0.0006, "step": 12300 }, { "epoch": 1.77, "learning_rate": 3.090592194887411e-05, "loss": 0.0012, "step": 12305 }, { "epoch": 1.78, "learning_rate": 3.0895905120602615e-05, "loss": 0.0007, "step": 12310 }, { "epoch": 1.78, "learning_rate": 3.088588829233112e-05, "loss": 0.0011, "step": 12315 }, { "epoch": 1.78, "learning_rate": 3.087587146405962e-05, "loss": 0.0009, "step": 12320 }, { "epoch": 1.78, "learning_rate": 3.086585463578813e-05, "loss": 0.0008, "step": 12325 }, { "epoch": 1.78, "learning_rate": 3.085583780751663e-05, "loss": 0.0007, "step": 12330 }, { "epoch": 1.78, "learning_rate": 3.084582097924513e-05, "loss": 0.0008, "step": 12335 }, { "epoch": 1.78, "learning_rate": 3.083580415097363e-05, "loss": 0.0011, "step": 12340 }, { "epoch": 1.78, "learning_rate": 3.082578732270214e-05, "loss": 0.0006, "step": 12345 }, { "epoch": 1.78, "learning_rate": 3.081577049443064e-05, "loss": 0.0009, "step": 12350 }, { "epoch": 1.78, "learning_rate": 3.080575366615915e-05, "loss": 0.001, "step": 12355 }, { "epoch": 1.78, "learning_rate": 3.0795736837887654e-05, "loss": 0.0008, "step": 12360 }, { "epoch": 1.78, "learning_rate": 3.078572000961616e-05, "loss": 0.0009, "step": 12365 }, { "epoch": 1.78, "learning_rate": 3.0775703181344664e-05, "loss": 0.0007, "step": 12370 }, { "epoch": 1.78, "learning_rate": 3.0765686353073166e-05, "loss": 0.0014, "step": 12375 }, { "epoch": 1.79, "learning_rate": 3.075566952480167e-05, "loss": 0.0009, "step": 12380 }, { "epoch": 1.79, "learning_rate": 3.074565269653017e-05, "loss": 0.0005, "step": 12385 }, { "epoch": 1.79, "learning_rate": 3.073563586825868e-05, "loss": 0.0006, "step": 12390 }, { "epoch": 1.79, "learning_rate": 3.072561903998718e-05, "loss": 0.0007, "step": 12395 }, { "epoch": 1.79, "learning_rate": 3.071560221171568e-05, "loss": 0.0009, "step": 12400 }, { "epoch": 1.79, "learning_rate": 3.0705585383444184e-05, "loss": 0.001, "step": 12405 }, { "epoch": 1.79, "learning_rate": 3.069556855517269e-05, "loss": 0.0011, "step": 12410 }, { "epoch": 1.79, "learning_rate": 3.0685551726901194e-05, "loss": 0.001, "step": 12415 }, { "epoch": 1.79, "learning_rate": 3.0675534898629696e-05, "loss": 0.001, "step": 12420 }, { "epoch": 1.79, "learning_rate": 3.0665518070358205e-05, "loss": 0.0011, "step": 12425 }, { "epoch": 1.79, "learning_rate": 3.0655501242086706e-05, "loss": 0.0008, "step": 12430 }, { "epoch": 1.79, "learning_rate": 3.0645484413815215e-05, "loss": 0.0006, "step": 12435 }, { "epoch": 1.79, "learning_rate": 3.063546758554372e-05, "loss": 0.0009, "step": 12440 }, { "epoch": 1.8, "learning_rate": 3.062545075727222e-05, "loss": 0.001, "step": 12445 }, { "epoch": 1.8, "learning_rate": 3.061543392900073e-05, "loss": 0.001, "step": 12450 }, { "epoch": 1.8, "learning_rate": 3.060541710072923e-05, "loss": 0.001, "step": 12455 }, { "epoch": 1.8, "learning_rate": 3.059540027245773e-05, "loss": 0.0011, "step": 12460 }, { "epoch": 1.8, "learning_rate": 3.058538344418623e-05, "loss": 0.0009, "step": 12465 }, { "epoch": 1.8, "learning_rate": 3.0575366615914735e-05, "loss": 0.0008, "step": 12470 }, { "epoch": 1.8, "learning_rate": 3.056534978764324e-05, "loss": 0.0009, "step": 12475 }, { "epoch": 1.8, "learning_rate": 3.0555332959371745e-05, "loss": 0.001, "step": 12480 }, { "epoch": 1.8, "learning_rate": 3.054531613110025e-05, "loss": 0.0007, "step": 12485 }, { "epoch": 1.8, "learning_rate": 3.053529930282875e-05, "loss": 0.0011, "step": 12490 }, { "epoch": 1.8, "learning_rate": 3.052528247455726e-05, "loss": 0.0011, "step": 12495 }, { "epoch": 1.8, "learning_rate": 3.0515265646285763e-05, "loss": 0.0008, "step": 12500 }, { "epoch": 1.8, "learning_rate": 3.0505248818014264e-05, "loss": 0.0009, "step": 12505 }, { "epoch": 1.8, "learning_rate": 3.0495231989742766e-05, "loss": 0.0014, "step": 12510 }, { "epoch": 1.81, "learning_rate": 3.0485215161471275e-05, "loss": 0.0008, "step": 12515 }, { "epoch": 1.81, "learning_rate": 3.0475198333199777e-05, "loss": 0.0009, "step": 12520 }, { "epoch": 1.81, "learning_rate": 3.0465181504928282e-05, "loss": 0.0007, "step": 12525 }, { "epoch": 1.81, "learning_rate": 3.0455164676656784e-05, "loss": 0.0009, "step": 12530 }, { "epoch": 1.81, "learning_rate": 3.0445147848385292e-05, "loss": 0.001, "step": 12535 }, { "epoch": 1.81, "learning_rate": 3.0435131020113794e-05, "loss": 0.0009, "step": 12540 }, { "epoch": 1.81, "learning_rate": 3.0425114191842296e-05, "loss": 0.0007, "step": 12545 }, { "epoch": 1.81, "learning_rate": 3.0415097363570798e-05, "loss": 0.0008, "step": 12550 }, { "epoch": 1.81, "learning_rate": 3.0405080535299303e-05, "loss": 0.0012, "step": 12555 }, { "epoch": 1.81, "learning_rate": 3.039506370702781e-05, "loss": 0.001, "step": 12560 }, { "epoch": 1.81, "learning_rate": 3.0385046878756314e-05, "loss": 0.0009, "step": 12565 }, { "epoch": 1.81, "learning_rate": 3.0375030050484815e-05, "loss": 0.0006, "step": 12570 }, { "epoch": 1.81, "learning_rate": 3.0365013222213317e-05, "loss": 0.001, "step": 12575 }, { "epoch": 1.81, "learning_rate": 3.0354996393941826e-05, "loss": 0.001, "step": 12580 }, { "epoch": 1.82, "learning_rate": 3.0344979565670328e-05, "loss": 0.0008, "step": 12585 }, { "epoch": 1.82, "learning_rate": 3.033496273739883e-05, "loss": 0.0005, "step": 12590 }, { "epoch": 1.82, "learning_rate": 3.0324945909127335e-05, "loss": 0.001, "step": 12595 }, { "epoch": 1.82, "learning_rate": 3.031492908085584e-05, "loss": 0.0004, "step": 12600 }, { "epoch": 1.82, "learning_rate": 3.0304912252584345e-05, "loss": 0.0011, "step": 12605 }, { "epoch": 1.82, "learning_rate": 3.0294895424312847e-05, "loss": 0.0008, "step": 12610 }, { "epoch": 1.82, "learning_rate": 3.028487859604135e-05, "loss": 0.001, "step": 12615 }, { "epoch": 1.82, "learning_rate": 3.027486176776985e-05, "loss": 0.0009, "step": 12620 }, { "epoch": 1.82, "learning_rate": 3.026484493949836e-05, "loss": 0.0008, "step": 12625 }, { "epoch": 1.82, "learning_rate": 3.025482811122686e-05, "loss": 0.0008, "step": 12630 }, { "epoch": 1.82, "learning_rate": 3.0244811282955366e-05, "loss": 0.0008, "step": 12635 }, { "epoch": 1.82, "learning_rate": 3.023479445468387e-05, "loss": 0.0009, "step": 12640 }, { "epoch": 1.82, "learning_rate": 3.0224777626412377e-05, "loss": 0.0011, "step": 12645 }, { "epoch": 1.82, "learning_rate": 3.021476079814088e-05, "loss": 0.0011, "step": 12650 }, { "epoch": 1.83, "learning_rate": 3.020474396986938e-05, "loss": 0.0007, "step": 12655 }, { "epoch": 1.83, "learning_rate": 3.0194727141597882e-05, "loss": 0.0017, "step": 12660 }, { "epoch": 1.83, "learning_rate": 3.018471031332639e-05, "loss": 0.0009, "step": 12665 }, { "epoch": 1.83, "learning_rate": 3.0174693485054893e-05, "loss": 0.0009, "step": 12670 }, { "epoch": 1.83, "learning_rate": 3.0164676656783398e-05, "loss": 0.0009, "step": 12675 }, { "epoch": 1.83, "learning_rate": 3.01546598285119e-05, "loss": 0.0009, "step": 12680 }, { "epoch": 1.83, "learning_rate": 3.014464300024041e-05, "loss": 0.0011, "step": 12685 }, { "epoch": 1.83, "learning_rate": 3.013462617196891e-05, "loss": 0.0011, "step": 12690 }, { "epoch": 1.83, "learning_rate": 3.0124609343697412e-05, "loss": 0.0009, "step": 12695 }, { "epoch": 1.83, "learning_rate": 3.0114592515425914e-05, "loss": 0.0012, "step": 12700 }, { "epoch": 1.83, "learning_rate": 3.010457568715442e-05, "loss": 0.0006, "step": 12705 }, { "epoch": 1.83, "learning_rate": 3.0094558858882928e-05, "loss": 0.0005, "step": 12710 }, { "epoch": 1.83, "learning_rate": 3.008454203061143e-05, "loss": 0.0011, "step": 12715 }, { "epoch": 1.83, "learning_rate": 3.007452520233993e-05, "loss": 0.0008, "step": 12720 }, { "epoch": 1.84, "learning_rate": 3.0064508374068433e-05, "loss": 0.0009, "step": 12725 }, { "epoch": 1.84, "learning_rate": 3.0054491545796942e-05, "loss": 0.0009, "step": 12730 }, { "epoch": 1.84, "learning_rate": 3.0044474717525444e-05, "loss": 0.0012, "step": 12735 }, { "epoch": 1.84, "learning_rate": 3.0034457889253946e-05, "loss": 0.0008, "step": 12740 }, { "epoch": 1.84, "learning_rate": 3.002444106098245e-05, "loss": 0.0013, "step": 12745 }, { "epoch": 1.84, "learning_rate": 3.001442423271096e-05, "loss": 0.0007, "step": 12750 }, { "epoch": 1.84, "learning_rate": 3.000440740443946e-05, "loss": 0.0011, "step": 12755 }, { "epoch": 1.84, "learning_rate": 2.9994390576167963e-05, "loss": 0.0007, "step": 12760 }, { "epoch": 1.84, "learning_rate": 2.9984373747896465e-05, "loss": 0.001, "step": 12765 }, { "epoch": 1.84, "learning_rate": 2.9974356919624974e-05, "loss": 0.0006, "step": 12770 }, { "epoch": 1.84, "learning_rate": 2.9964340091353476e-05, "loss": 0.0007, "step": 12775 }, { "epoch": 1.84, "learning_rate": 2.995432326308198e-05, "loss": 0.001, "step": 12780 }, { "epoch": 1.84, "learning_rate": 2.9944306434810483e-05, "loss": 0.0006, "step": 12785 }, { "epoch": 1.84, "learning_rate": 2.9934289606538984e-05, "loss": 0.0007, "step": 12790 }, { "epoch": 1.85, "learning_rate": 2.9924272778267493e-05, "loss": 0.001, "step": 12795 }, { "epoch": 1.85, "learning_rate": 2.9914255949995995e-05, "loss": 0.0008, "step": 12800 }, { "epoch": 1.85, "learning_rate": 2.9904239121724497e-05, "loss": 0.0013, "step": 12805 }, { "epoch": 1.85, "learning_rate": 2.9894222293453e-05, "loss": 0.001, "step": 12810 }, { "epoch": 1.85, "learning_rate": 2.9884205465181507e-05, "loss": 0.0008, "step": 12815 }, { "epoch": 1.85, "learning_rate": 2.9874188636910012e-05, "loss": 0.0016, "step": 12820 }, { "epoch": 1.85, "learning_rate": 2.9864171808638514e-05, "loss": 0.0006, "step": 12825 }, { "epoch": 1.85, "learning_rate": 2.9854154980367016e-05, "loss": 0.0012, "step": 12830 }, { "epoch": 1.85, "learning_rate": 2.9844138152095525e-05, "loss": 0.0011, "step": 12835 }, { "epoch": 1.85, "learning_rate": 2.9834121323824027e-05, "loss": 0.0012, "step": 12840 }, { "epoch": 1.85, "learning_rate": 2.982410449555253e-05, "loss": 0.0012, "step": 12845 }, { "epoch": 1.85, "learning_rate": 2.9814087667281034e-05, "loss": 0.0009, "step": 12850 }, { "epoch": 1.85, "learning_rate": 2.980407083900954e-05, "loss": 0.0013, "step": 12855 }, { "epoch": 1.85, "learning_rate": 2.9794054010738044e-05, "loss": 0.0009, "step": 12860 }, { "epoch": 1.86, "learning_rate": 2.9784037182466546e-05, "loss": 0.0008, "step": 12865 }, { "epoch": 1.86, "learning_rate": 2.9774020354195048e-05, "loss": 0.0009, "step": 12870 }, { "epoch": 1.86, "learning_rate": 2.976400352592355e-05, "loss": 0.0008, "step": 12875 }, { "epoch": 1.86, "learning_rate": 2.9753986697652058e-05, "loss": 0.0009, "step": 12880 }, { "epoch": 1.86, "learning_rate": 2.974396986938056e-05, "loss": 0.0007, "step": 12885 }, { "epoch": 1.86, "learning_rate": 2.9733953041109065e-05, "loss": 0.0014, "step": 12890 }, { "epoch": 1.86, "learning_rate": 2.9723936212837567e-05, "loss": 0.0005, "step": 12895 }, { "epoch": 1.86, "learning_rate": 2.9713919384566076e-05, "loss": 0.001, "step": 12900 }, { "epoch": 1.86, "learning_rate": 2.9703902556294578e-05, "loss": 0.001, "step": 12905 }, { "epoch": 1.86, "learning_rate": 2.969388572802308e-05, "loss": 0.0009, "step": 12910 }, { "epoch": 1.86, "learning_rate": 2.968386889975158e-05, "loss": 0.0009, "step": 12915 }, { "epoch": 1.86, "learning_rate": 2.967385207148009e-05, "loss": 0.0009, "step": 12920 }, { "epoch": 1.86, "learning_rate": 2.9663835243208592e-05, "loss": 0.0008, "step": 12925 }, { "epoch": 1.86, "learning_rate": 2.9653818414937097e-05, "loss": 0.0011, "step": 12930 }, { "epoch": 1.87, "learning_rate": 2.96438015866656e-05, "loss": 0.0008, "step": 12935 }, { "epoch": 1.87, "learning_rate": 2.96337847583941e-05, "loss": 0.0011, "step": 12940 }, { "epoch": 1.87, "learning_rate": 2.962376793012261e-05, "loss": 0.0008, "step": 12945 }, { "epoch": 1.87, "learning_rate": 2.961375110185111e-05, "loss": 0.001, "step": 12950 }, { "epoch": 1.87, "learning_rate": 2.9603734273579613e-05, "loss": 0.0009, "step": 12955 }, { "epoch": 1.87, "learning_rate": 2.9593717445308118e-05, "loss": 0.0008, "step": 12960 }, { "epoch": 1.87, "learning_rate": 2.9583700617036623e-05, "loss": 0.0009, "step": 12965 }, { "epoch": 1.87, "learning_rate": 2.957368378876513e-05, "loss": 0.0006, "step": 12970 }, { "epoch": 1.87, "learning_rate": 2.956366696049363e-05, "loss": 0.0008, "step": 12975 }, { "epoch": 1.87, "learning_rate": 2.9553650132222132e-05, "loss": 0.0008, "step": 12980 }, { "epoch": 1.87, "learning_rate": 2.954363330395064e-05, "loss": 0.0011, "step": 12985 }, { "epoch": 1.87, "learning_rate": 2.9533616475679143e-05, "loss": 0.0009, "step": 12990 }, { "epoch": 1.87, "learning_rate": 2.9523599647407645e-05, "loss": 0.0009, "step": 12995 }, { "epoch": 1.88, "learning_rate": 2.951358281913615e-05, "loss": 0.001, "step": 13000 }, { "epoch": 1.88, "learning_rate": 2.9503565990864655e-05, "loss": 0.0009, "step": 13005 }, { "epoch": 1.88, "learning_rate": 2.949354916259316e-05, "loss": 0.001, "step": 13010 }, { "epoch": 1.88, "learning_rate": 2.9483532334321662e-05, "loss": 0.0006, "step": 13015 }, { "epoch": 1.88, "learning_rate": 2.9473515506050164e-05, "loss": 0.0011, "step": 13020 }, { "epoch": 1.88, "learning_rate": 2.9463498677778666e-05, "loss": 0.0011, "step": 13025 }, { "epoch": 1.88, "learning_rate": 2.9453481849507174e-05, "loss": 0.0009, "step": 13030 }, { "epoch": 1.88, "learning_rate": 2.9443465021235676e-05, "loss": 0.0012, "step": 13035 }, { "epoch": 1.88, "learning_rate": 2.943344819296418e-05, "loss": 0.0014, "step": 13040 }, { "epoch": 1.88, "learning_rate": 2.9423431364692683e-05, "loss": 0.001, "step": 13045 }, { "epoch": 1.88, "learning_rate": 2.9413414536421192e-05, "loss": 0.0011, "step": 13050 }, { "epoch": 1.88, "learning_rate": 2.9403397708149694e-05, "loss": 0.0015, "step": 13055 }, { "epoch": 1.88, "learning_rate": 2.9393380879878196e-05, "loss": 0.0014, "step": 13060 }, { "epoch": 1.88, "learning_rate": 2.9383364051606697e-05, "loss": 0.0006, "step": 13065 }, { "epoch": 1.89, "learning_rate": 2.9373347223335206e-05, "loss": 0.0013, "step": 13070 }, { "epoch": 1.89, "learning_rate": 2.9363330395063708e-05, "loss": 0.0008, "step": 13075 }, { "epoch": 1.89, "learning_rate": 2.9353313566792213e-05, "loss": 0.0008, "step": 13080 }, { "epoch": 1.89, "learning_rate": 2.9343296738520715e-05, "loss": 0.0008, "step": 13085 }, { "epoch": 1.89, "learning_rate": 2.9333279910249224e-05, "loss": 0.001, "step": 13090 }, { "epoch": 1.89, "learning_rate": 2.9323263081977725e-05, "loss": 0.0012, "step": 13095 }, { "epoch": 1.89, "learning_rate": 2.9313246253706227e-05, "loss": 0.0011, "step": 13100 }, { "epoch": 1.89, "learning_rate": 2.930322942543473e-05, "loss": 0.0007, "step": 13105 }, { "epoch": 1.89, "learning_rate": 2.9293212597163234e-05, "loss": 0.001, "step": 13110 }, { "epoch": 1.89, "learning_rate": 2.9283195768891743e-05, "loss": 0.0009, "step": 13115 }, { "epoch": 1.89, "learning_rate": 2.9273178940620245e-05, "loss": 0.0013, "step": 13120 }, { "epoch": 1.89, "learning_rate": 2.9263162112348747e-05, "loss": 0.0015, "step": 13125 }, { "epoch": 1.89, "learning_rate": 2.925314528407725e-05, "loss": 0.0013, "step": 13130 }, { "epoch": 1.89, "learning_rate": 2.9243128455805757e-05, "loss": 0.0011, "step": 13135 }, { "epoch": 1.9, "learning_rate": 2.923311162753426e-05, "loss": 0.0011, "step": 13140 }, { "epoch": 1.9, "learning_rate": 2.922309479926276e-05, "loss": 0.0006, "step": 13145 }, { "epoch": 1.9, "learning_rate": 2.9213077970991266e-05, "loss": 0.0014, "step": 13150 }, { "epoch": 1.9, "learning_rate": 2.9203061142719775e-05, "loss": 0.0008, "step": 13155 }, { "epoch": 1.9, "learning_rate": 2.9193044314448276e-05, "loss": 0.0012, "step": 13160 }, { "epoch": 1.9, "learning_rate": 2.9183027486176778e-05, "loss": 0.0008, "step": 13165 }, { "epoch": 1.9, "learning_rate": 2.917301065790528e-05, "loss": 0.001, "step": 13170 }, { "epoch": 1.9, "learning_rate": 2.9162993829633782e-05, "loss": 0.0009, "step": 13175 }, { "epoch": 1.9, "learning_rate": 2.915297700136229e-05, "loss": 0.001, "step": 13180 }, { "epoch": 1.9, "learning_rate": 2.9142960173090796e-05, "loss": 0.0009, "step": 13185 }, { "epoch": 1.9, "learning_rate": 2.9132943344819298e-05, "loss": 0.0009, "step": 13190 }, { "epoch": 1.9, "learning_rate": 2.91229265165478e-05, "loss": 0.0007, "step": 13195 }, { "epoch": 1.9, "learning_rate": 2.9112909688276308e-05, "loss": 0.0009, "step": 13200 }, { "epoch": 1.9, "learning_rate": 2.910289286000481e-05, "loss": 0.0007, "step": 13205 }, { "epoch": 1.91, "learning_rate": 2.9092876031733312e-05, "loss": 0.0012, "step": 13210 }, { "epoch": 1.91, "learning_rate": 2.9082859203461817e-05, "loss": 0.0007, "step": 13215 }, { "epoch": 1.91, "learning_rate": 2.9072842375190322e-05, "loss": 0.0007, "step": 13220 }, { "epoch": 1.91, "learning_rate": 2.9062825546918827e-05, "loss": 0.0014, "step": 13225 }, { "epoch": 1.91, "learning_rate": 2.905280871864733e-05, "loss": 0.001, "step": 13230 }, { "epoch": 1.91, "learning_rate": 2.904279189037583e-05, "loss": 0.0008, "step": 13235 }, { "epoch": 1.91, "learning_rate": 2.903277506210434e-05, "loss": 0.001, "step": 13240 }, { "epoch": 1.91, "learning_rate": 2.902275823383284e-05, "loss": 0.0012, "step": 13245 }, { "epoch": 1.91, "learning_rate": 2.9012741405561343e-05, "loss": 0.0008, "step": 13250 }, { "epoch": 1.91, "learning_rate": 2.900272457728985e-05, "loss": 0.0011, "step": 13255 }, { "epoch": 1.91, "learning_rate": 2.899270774901835e-05, "loss": 0.0006, "step": 13260 }, { "epoch": 1.91, "learning_rate": 2.898269092074686e-05, "loss": 0.001, "step": 13265 }, { "epoch": 1.91, "learning_rate": 2.897267409247536e-05, "loss": 0.0012, "step": 13270 }, { "epoch": 1.91, "learning_rate": 2.8962657264203863e-05, "loss": 0.0009, "step": 13275 }, { "epoch": 1.92, "learning_rate": 2.8952640435932365e-05, "loss": 0.0007, "step": 13280 }, { "epoch": 1.92, "learning_rate": 2.8942623607660873e-05, "loss": 0.0008, "step": 13285 }, { "epoch": 1.92, "learning_rate": 2.8932606779389375e-05, "loss": 0.0009, "step": 13290 }, { "epoch": 1.92, "learning_rate": 2.892258995111788e-05, "loss": 0.001, "step": 13295 }, { "epoch": 1.92, "learning_rate": 2.8912573122846382e-05, "loss": 0.0009, "step": 13300 }, { "epoch": 1.92, "learning_rate": 2.890255629457489e-05, "loss": 0.0005, "step": 13305 }, { "epoch": 1.92, "learning_rate": 2.8892539466303393e-05, "loss": 0.0009, "step": 13310 }, { "epoch": 1.92, "learning_rate": 2.8882522638031894e-05, "loss": 0.0007, "step": 13315 }, { "epoch": 1.92, "learning_rate": 2.8872505809760396e-05, "loss": 0.0008, "step": 13320 }, { "epoch": 1.92, "learning_rate": 2.8862488981488905e-05, "loss": 0.0007, "step": 13325 }, { "epoch": 1.92, "learning_rate": 2.8852472153217407e-05, "loss": 0.001, "step": 13330 }, { "epoch": 1.92, "learning_rate": 2.8842455324945912e-05, "loss": 0.0007, "step": 13335 }, { "epoch": 1.92, "learning_rate": 2.8832438496674414e-05, "loss": 0.0011, "step": 13340 }, { "epoch": 1.92, "learning_rate": 2.8822421668402916e-05, "loss": 0.0009, "step": 13345 }, { "epoch": 1.93, "learning_rate": 2.8812404840131424e-05, "loss": 0.0007, "step": 13350 }, { "epoch": 1.93, "learning_rate": 2.8802388011859926e-05, "loss": 0.0009, "step": 13355 }, { "epoch": 1.93, "learning_rate": 2.8792371183588428e-05, "loss": 0.001, "step": 13360 }, { "epoch": 1.93, "learning_rate": 2.8782354355316933e-05, "loss": 0.0008, "step": 13365 }, { "epoch": 1.93, "learning_rate": 2.877233752704544e-05, "loss": 0.001, "step": 13370 }, { "epoch": 1.93, "learning_rate": 2.8762320698773944e-05, "loss": 0.001, "step": 13375 }, { "epoch": 1.93, "learning_rate": 2.8752303870502445e-05, "loss": 0.0012, "step": 13380 }, { "epoch": 1.93, "learning_rate": 2.8742287042230947e-05, "loss": 0.0008, "step": 13385 }, { "epoch": 1.93, "learning_rate": 2.8732270213959456e-05, "loss": 0.001, "step": 13390 }, { "epoch": 1.93, "learning_rate": 2.8722253385687958e-05, "loss": 0.001, "step": 13395 }, { "epoch": 1.93, "learning_rate": 2.871223655741646e-05, "loss": 0.0012, "step": 13400 }, { "epoch": 1.93, "learning_rate": 2.8702219729144965e-05, "loss": 0.001, "step": 13405 }, { "epoch": 1.93, "learning_rate": 2.8692202900873467e-05, "loss": 0.0011, "step": 13410 }, { "epoch": 1.93, "learning_rate": 2.8682186072601975e-05, "loss": 0.0012, "step": 13415 }, { "epoch": 1.94, "learning_rate": 2.8672169244330477e-05, "loss": 0.0008, "step": 13420 }, { "epoch": 1.94, "learning_rate": 2.866215241605898e-05, "loss": 0.0012, "step": 13425 }, { "epoch": 1.94, "learning_rate": 2.865213558778748e-05, "loss": 0.0009, "step": 13430 }, { "epoch": 1.94, "learning_rate": 2.864211875951599e-05, "loss": 0.0008, "step": 13435 }, { "epoch": 1.94, "learning_rate": 2.863210193124449e-05, "loss": 0.0011, "step": 13440 }, { "epoch": 1.94, "learning_rate": 2.8622085102972996e-05, "loss": 0.001, "step": 13445 }, { "epoch": 1.94, "learning_rate": 2.8612068274701498e-05, "loss": 0.0013, "step": 13450 }, { "epoch": 1.94, "learning_rate": 2.8602051446430007e-05, "loss": 0.0011, "step": 13455 }, { "epoch": 1.94, "learning_rate": 2.859203461815851e-05, "loss": 0.0009, "step": 13460 }, { "epoch": 1.94, "learning_rate": 2.858201778988701e-05, "loss": 0.0007, "step": 13465 }, { "epoch": 1.94, "learning_rate": 2.8572000961615512e-05, "loss": 0.001, "step": 13470 }, { "epoch": 1.94, "learning_rate": 2.856198413334402e-05, "loss": 0.0007, "step": 13475 }, { "epoch": 1.94, "learning_rate": 2.8551967305072523e-05, "loss": 0.0008, "step": 13480 }, { "epoch": 1.95, "learning_rate": 2.8541950476801028e-05, "loss": 0.001, "step": 13485 }, { "epoch": 1.95, "learning_rate": 2.853193364852953e-05, "loss": 0.0007, "step": 13490 }, { "epoch": 1.95, "learning_rate": 2.8521916820258032e-05, "loss": 0.0007, "step": 13495 }, { "epoch": 1.95, "learning_rate": 2.851189999198654e-05, "loss": 0.0006, "step": 13500 }, { "epoch": 1.95, "learning_rate": 2.8501883163715042e-05, "loss": 0.0007, "step": 13505 }, { "epoch": 1.95, "learning_rate": 2.8491866335443544e-05, "loss": 0.001, "step": 13510 }, { "epoch": 1.95, "learning_rate": 2.848184950717205e-05, "loss": 0.0007, "step": 13515 }, { "epoch": 1.95, "learning_rate": 2.8471832678900558e-05, "loss": 0.0004, "step": 13520 }, { "epoch": 1.95, "learning_rate": 2.846181585062906e-05, "loss": 0.0014, "step": 13525 }, { "epoch": 1.95, "learning_rate": 2.845179902235756e-05, "loss": 0.0009, "step": 13530 }, { "epoch": 1.95, "learning_rate": 2.8441782194086063e-05, "loss": 0.001, "step": 13535 }, { "epoch": 1.95, "learning_rate": 2.8431765365814572e-05, "loss": 0.0009, "step": 13540 }, { "epoch": 1.95, "learning_rate": 2.8421748537543074e-05, "loss": 0.0011, "step": 13545 }, { "epoch": 1.95, "learning_rate": 2.8411731709271576e-05, "loss": 0.0009, "step": 13550 }, { "epoch": 1.96, "learning_rate": 2.840171488100008e-05, "loss": 0.001, "step": 13555 }, { "epoch": 1.96, "learning_rate": 2.839169805272859e-05, "loss": 0.0012, "step": 13560 }, { "epoch": 1.96, "learning_rate": 2.838168122445709e-05, "loss": 0.001, "step": 13565 }, { "epoch": 1.96, "learning_rate": 2.8371664396185593e-05, "loss": 0.0012, "step": 13570 }, { "epoch": 1.96, "learning_rate": 2.8361647567914095e-05, "loss": 0.0007, "step": 13575 }, { "epoch": 1.96, "learning_rate": 2.8351630739642597e-05, "loss": 0.001, "step": 13580 }, { "epoch": 1.96, "learning_rate": 2.8341613911371105e-05, "loss": 0.0011, "step": 13585 }, { "epoch": 1.96, "learning_rate": 2.833159708309961e-05, "loss": 0.0012, "step": 13590 }, { "epoch": 1.96, "learning_rate": 2.8321580254828113e-05, "loss": 0.001, "step": 13595 }, { "epoch": 1.96, "learning_rate": 2.8311563426556614e-05, "loss": 0.0008, "step": 13600 }, { "epoch": 1.96, "learning_rate": 2.8301546598285123e-05, "loss": 0.0013, "step": 13605 }, { "epoch": 1.96, "learning_rate": 2.8291529770013625e-05, "loss": 0.001, "step": 13610 }, { "epoch": 1.96, "learning_rate": 2.8281512941742127e-05, "loss": 0.0011, "step": 13615 }, { "epoch": 1.96, "learning_rate": 2.8271496113470632e-05, "loss": 0.0008, "step": 13620 }, { "epoch": 1.97, "learning_rate": 2.8261479285199137e-05, "loss": 0.0006, "step": 13625 }, { "epoch": 1.97, "learning_rate": 2.8251462456927642e-05, "loss": 0.0005, "step": 13630 }, { "epoch": 1.97, "learning_rate": 2.8241445628656144e-05, "loss": 0.001, "step": 13635 }, { "epoch": 1.97, "learning_rate": 2.8231428800384646e-05, "loss": 0.0011, "step": 13640 }, { "epoch": 1.97, "learning_rate": 2.8221411972113148e-05, "loss": 0.0007, "step": 13645 }, { "epoch": 1.97, "learning_rate": 2.8211395143841656e-05, "loss": 0.0008, "step": 13650 }, { "epoch": 1.97, "learning_rate": 2.820137831557016e-05, "loss": 0.0006, "step": 13655 }, { "epoch": 1.97, "learning_rate": 2.8191361487298664e-05, "loss": 0.0008, "step": 13660 }, { "epoch": 1.97, "learning_rate": 2.8181344659027165e-05, "loss": 0.001, "step": 13665 }, { "epoch": 1.97, "learning_rate": 2.8171327830755674e-05, "loss": 0.001, "step": 13670 }, { "epoch": 1.97, "learning_rate": 2.8161311002484176e-05, "loss": 0.0009, "step": 13675 }, { "epoch": 1.97, "learning_rate": 2.8151294174212678e-05, "loss": 0.0011, "step": 13680 }, { "epoch": 1.97, "learning_rate": 2.814127734594118e-05, "loss": 0.0008, "step": 13685 }, { "epoch": 1.97, "learning_rate": 2.8131260517669688e-05, "loss": 0.0008, "step": 13690 }, { "epoch": 1.98, "learning_rate": 2.812124368939819e-05, "loss": 0.001, "step": 13695 }, { "epoch": 1.98, "learning_rate": 2.8111226861126695e-05, "loss": 0.0008, "step": 13700 }, { "epoch": 1.98, "learning_rate": 2.8101210032855197e-05, "loss": 0.0009, "step": 13705 }, { "epoch": 1.98, "learning_rate": 2.8091193204583706e-05, "loss": 0.0007, "step": 13710 }, { "epoch": 1.98, "learning_rate": 2.8081176376312207e-05, "loss": 0.0013, "step": 13715 }, { "epoch": 1.98, "learning_rate": 2.807115954804071e-05, "loss": 0.0013, "step": 13720 }, { "epoch": 1.98, "learning_rate": 2.806114271976921e-05, "loss": 0.0009, "step": 13725 }, { "epoch": 1.98, "learning_rate": 2.8051125891497716e-05, "loss": 0.0014, "step": 13730 }, { "epoch": 1.98, "learning_rate": 2.804110906322622e-05, "loss": 0.0009, "step": 13735 }, { "epoch": 1.98, "learning_rate": 2.8031092234954727e-05, "loss": 0.0007, "step": 13740 }, { "epoch": 1.98, "learning_rate": 2.802107540668323e-05, "loss": 0.0009, "step": 13745 }, { "epoch": 1.98, "learning_rate": 2.801105857841173e-05, "loss": 0.0008, "step": 13750 }, { "epoch": 1.98, "learning_rate": 2.800104175014024e-05, "loss": 0.001, "step": 13755 }, { "epoch": 1.98, "learning_rate": 2.799102492186874e-05, "loss": 0.0009, "step": 13760 }, { "epoch": 1.99, "learning_rate": 2.7981008093597243e-05, "loss": 0.0008, "step": 13765 }, { "epoch": 1.99, "learning_rate": 2.7970991265325748e-05, "loss": 0.001, "step": 13770 }, { "epoch": 1.99, "learning_rate": 2.7960974437054253e-05, "loss": 0.001, "step": 13775 }, { "epoch": 1.99, "learning_rate": 2.795095760878276e-05, "loss": 0.0008, "step": 13780 }, { "epoch": 1.99, "learning_rate": 2.794094078051126e-05, "loss": 0.0011, "step": 13785 }, { "epoch": 1.99, "learning_rate": 2.7930923952239762e-05, "loss": 0.0013, "step": 13790 }, { "epoch": 1.99, "learning_rate": 2.792090712396827e-05, "loss": 0.0011, "step": 13795 }, { "epoch": 1.99, "learning_rate": 2.7910890295696773e-05, "loss": 0.0009, "step": 13800 }, { "epoch": 1.99, "learning_rate": 2.7900873467425274e-05, "loss": 0.0007, "step": 13805 }, { "epoch": 1.99, "learning_rate": 2.789085663915378e-05, "loss": 0.0005, "step": 13810 }, { "epoch": 1.99, "learning_rate": 2.788083981088228e-05, "loss": 0.0008, "step": 13815 }, { "epoch": 1.99, "learning_rate": 2.787082298261079e-05, "loss": 0.0008, "step": 13820 }, { "epoch": 1.99, "learning_rate": 2.7860806154339292e-05, "loss": 0.0011, "step": 13825 }, { "epoch": 1.99, "learning_rate": 2.7850789326067794e-05, "loss": 0.0011, "step": 13830 }, { "epoch": 2.0, "learning_rate": 2.7840772497796296e-05, "loss": 0.0008, "step": 13835 }, { "epoch": 2.0, "learning_rate": 2.7830755669524804e-05, "loss": 0.0008, "step": 13840 }, { "epoch": 2.0, "learning_rate": 2.7820738841253306e-05, "loss": 0.0014, "step": 13845 }, { "epoch": 2.0, "learning_rate": 2.781072201298181e-05, "loss": 0.0008, "step": 13850 }, { "epoch": 2.0, "learning_rate": 2.7800705184710313e-05, "loss": 0.0011, "step": 13855 }, { "epoch": 2.0, "learning_rate": 2.7790688356438822e-05, "loss": 0.0007, "step": 13860 }, { "epoch": 2.0, "learning_rate": 2.7780671528167324e-05, "loss": 0.0012, "step": 13865 }, { "epoch": 2.0, "learning_rate": 2.7770654699895825e-05, "loss": 0.0006, "step": 13870 }, { "epoch": 2.0, "learning_rate": 2.7760637871624327e-05, "loss": 0.0012, "step": 13875 }, { "epoch": 2.0, "learning_rate": 2.7750621043352836e-05, "loss": 0.0012, "step": 13880 }, { "epoch": 2.0, "learning_rate": 2.7740604215081338e-05, "loss": 0.0011, "step": 13885 }, { "epoch": 2.0, "learning_rate": 2.7730587386809843e-05, "loss": 0.0008, "step": 13890 }, { "epoch": 2.0, "learning_rate": 2.7720570558538345e-05, "loss": 0.0008, "step": 13895 }, { "epoch": 2.0, "learning_rate": 2.7710553730266847e-05, "loss": 0.0009, "step": 13900 }, { "epoch": 2.01, "learning_rate": 2.7700536901995355e-05, "loss": 0.0007, "step": 13905 }, { "epoch": 2.01, "learning_rate": 2.7690520073723857e-05, "loss": 0.001, "step": 13910 }, { "epoch": 2.01, "learning_rate": 2.768050324545236e-05, "loss": 0.0007, "step": 13915 }, { "epoch": 2.01, "learning_rate": 2.7670486417180864e-05, "loss": 0.0013, "step": 13920 }, { "epoch": 2.01, "learning_rate": 2.7660469588909373e-05, "loss": 0.0012, "step": 13925 }, { "epoch": 2.01, "learning_rate": 2.7650452760637875e-05, "loss": 0.0015, "step": 13930 }, { "epoch": 2.01, "learning_rate": 2.7640435932366376e-05, "loss": 0.0004, "step": 13935 }, { "epoch": 2.01, "learning_rate": 2.763041910409488e-05, "loss": 0.0006, "step": 13940 }, { "epoch": 2.01, "learning_rate": 2.7620402275823387e-05, "loss": 0.0009, "step": 13945 }, { "epoch": 2.01, "learning_rate": 2.761038544755189e-05, "loss": 0.0008, "step": 13950 }, { "epoch": 2.01, "learning_rate": 2.7600368619280394e-05, "loss": 0.0009, "step": 13955 }, { "epoch": 2.01, "learning_rate": 2.7590351791008896e-05, "loss": 0.0009, "step": 13960 }, { "epoch": 2.01, "learning_rate": 2.7580334962737398e-05, "loss": 0.0008, "step": 13965 }, { "epoch": 2.02, "learning_rate": 2.7570318134465906e-05, "loss": 0.0006, "step": 13970 }, { "epoch": 2.02, "learning_rate": 2.7560301306194408e-05, "loss": 0.0008, "step": 13975 }, { "epoch": 2.02, "learning_rate": 2.755028447792291e-05, "loss": 0.0005, "step": 13980 }, { "epoch": 2.02, "learning_rate": 2.7540267649651412e-05, "loss": 0.0009, "step": 13985 }, { "epoch": 2.02, "learning_rate": 2.753025082137992e-05, "loss": 0.0007, "step": 13990 }, { "epoch": 2.02, "learning_rate": 2.7520233993108426e-05, "loss": 0.0007, "step": 13995 }, { "epoch": 2.02, "learning_rate": 2.7510217164836927e-05, "loss": 0.0013, "step": 14000 }, { "epoch": 2.02, "learning_rate": 2.750020033656543e-05, "loss": 0.0009, "step": 14005 }, { "epoch": 2.02, "learning_rate": 2.7490183508293938e-05, "loss": 0.0009, "step": 14010 }, { "epoch": 2.02, "learning_rate": 2.748016668002244e-05, "loss": 0.0008, "step": 14015 }, { "epoch": 2.02, "learning_rate": 2.747014985175094e-05, "loss": 0.0011, "step": 14020 }, { "epoch": 2.02, "learning_rate": 2.7460133023479447e-05, "loss": 0.001, "step": 14025 }, { "epoch": 2.02, "learning_rate": 2.7450116195207952e-05, "loss": 0.0009, "step": 14030 }, { "epoch": 2.02, "learning_rate": 2.7440099366936457e-05, "loss": 0.0005, "step": 14035 }, { "epoch": 2.03, "learning_rate": 2.743008253866496e-05, "loss": 0.0008, "step": 14040 }, { "epoch": 2.03, "learning_rate": 2.742006571039346e-05, "loss": 0.0005, "step": 14045 }, { "epoch": 2.03, "learning_rate": 2.7410048882121963e-05, "loss": 0.0009, "step": 14050 }, { "epoch": 2.03, "learning_rate": 2.740003205385047e-05, "loss": 0.0005, "step": 14055 }, { "epoch": 2.03, "learning_rate": 2.7390015225578973e-05, "loss": 0.0008, "step": 14060 }, { "epoch": 2.03, "learning_rate": 2.737999839730748e-05, "loss": 0.0013, "step": 14065 }, { "epoch": 2.03, "learning_rate": 2.736998156903598e-05, "loss": 0.0008, "step": 14070 }, { "epoch": 2.03, "learning_rate": 2.735996474076449e-05, "loss": 0.0009, "step": 14075 }, { "epoch": 2.03, "learning_rate": 2.734994791249299e-05, "loss": 0.0009, "step": 14080 }, { "epoch": 2.03, "learning_rate": 2.7339931084221493e-05, "loss": 0.0006, "step": 14085 }, { "epoch": 2.03, "learning_rate": 2.7329914255949994e-05, "loss": 0.0007, "step": 14090 }, { "epoch": 2.03, "learning_rate": 2.7319897427678503e-05, "loss": 0.0012, "step": 14095 }, { "epoch": 2.03, "learning_rate": 2.7309880599407005e-05, "loss": 0.001, "step": 14100 }, { "epoch": 2.03, "learning_rate": 2.729986377113551e-05, "loss": 0.001, "step": 14105 }, { "epoch": 2.04, "learning_rate": 2.7289846942864012e-05, "loss": 0.0009, "step": 14110 }, { "epoch": 2.04, "learning_rate": 2.727983011459252e-05, "loss": 0.0009, "step": 14115 }, { "epoch": 2.04, "learning_rate": 2.7269813286321022e-05, "loss": 0.001, "step": 14120 }, { "epoch": 2.04, "learning_rate": 2.7259796458049524e-05, "loss": 0.0006, "step": 14125 }, { "epoch": 2.04, "learning_rate": 2.7249779629778026e-05, "loss": 0.0006, "step": 14130 }, { "epoch": 2.04, "learning_rate": 2.723976280150653e-05, "loss": 0.0009, "step": 14135 }, { "epoch": 2.04, "learning_rate": 2.7229745973235037e-05, "loss": 0.0007, "step": 14140 }, { "epoch": 2.04, "learning_rate": 2.7219729144963542e-05, "loss": 0.0007, "step": 14145 }, { "epoch": 2.04, "learning_rate": 2.7209712316692044e-05, "loss": 0.0008, "step": 14150 }, { "epoch": 2.04, "learning_rate": 2.7199695488420545e-05, "loss": 0.0007, "step": 14155 }, { "epoch": 2.04, "learning_rate": 2.7189678660149054e-05, "loss": 0.0007, "step": 14160 }, { "epoch": 2.04, "learning_rate": 2.7179661831877556e-05, "loss": 0.0005, "step": 14165 }, { "epoch": 2.04, "learning_rate": 2.7169645003606058e-05, "loss": 0.0011, "step": 14170 }, { "epoch": 2.04, "learning_rate": 2.7159628175334563e-05, "loss": 0.0008, "step": 14175 }, { "epoch": 2.05, "learning_rate": 2.7149611347063068e-05, "loss": 0.0011, "step": 14180 }, { "epoch": 2.05, "learning_rate": 2.7139594518791573e-05, "loss": 0.0009, "step": 14185 }, { "epoch": 2.05, "learning_rate": 2.7129577690520075e-05, "loss": 0.0009, "step": 14190 }, { "epoch": 2.05, "learning_rate": 2.7119560862248577e-05, "loss": 0.0008, "step": 14195 }, { "epoch": 2.05, "learning_rate": 2.710954403397708e-05, "loss": 0.0011, "step": 14200 }, { "epoch": 2.05, "learning_rate": 2.7099527205705588e-05, "loss": 0.001, "step": 14205 }, { "epoch": 2.05, "learning_rate": 2.708951037743409e-05, "loss": 0.0006, "step": 14210 }, { "epoch": 2.05, "learning_rate": 2.7079493549162595e-05, "loss": 0.0008, "step": 14215 }, { "epoch": 2.05, "learning_rate": 2.7069476720891096e-05, "loss": 0.001, "step": 14220 }, { "epoch": 2.05, "learning_rate": 2.7059459892619605e-05, "loss": 0.0006, "step": 14225 }, { "epoch": 2.05, "learning_rate": 2.7049443064348107e-05, "loss": 0.0006, "step": 14230 }, { "epoch": 2.05, "learning_rate": 2.703942623607661e-05, "loss": 0.0007, "step": 14235 }, { "epoch": 2.05, "learning_rate": 2.702940940780511e-05, "loss": 0.0006, "step": 14240 }, { "epoch": 2.05, "learning_rate": 2.701939257953362e-05, "loss": 0.0008, "step": 14245 }, { "epoch": 2.06, "learning_rate": 2.700937575126212e-05, "loss": 0.0009, "step": 14250 }, { "epoch": 2.06, "learning_rate": 2.6999358922990626e-05, "loss": 0.0009, "step": 14255 }, { "epoch": 2.06, "learning_rate": 2.6989342094719128e-05, "loss": 0.0009, "step": 14260 }, { "epoch": 2.06, "learning_rate": 2.6979325266447637e-05, "loss": 0.0008, "step": 14265 }, { "epoch": 2.06, "learning_rate": 2.696930843817614e-05, "loss": 0.0008, "step": 14270 }, { "epoch": 2.06, "learning_rate": 2.695929160990464e-05, "loss": 0.0012, "step": 14275 }, { "epoch": 2.06, "learning_rate": 2.6949274781633142e-05, "loss": 0.0006, "step": 14280 }, { "epoch": 2.06, "learning_rate": 2.6939257953361647e-05, "loss": 0.0007, "step": 14285 }, { "epoch": 2.06, "learning_rate": 2.6929241125090156e-05, "loss": 0.0005, "step": 14290 }, { "epoch": 2.06, "learning_rate": 2.6919224296818658e-05, "loss": 0.0007, "step": 14295 }, { "epoch": 2.06, "learning_rate": 2.690920746854716e-05, "loss": 0.0007, "step": 14300 }, { "epoch": 2.06, "learning_rate": 2.689919064027566e-05, "loss": 0.0011, "step": 14305 }, { "epoch": 2.06, "learning_rate": 2.688917381200417e-05, "loss": 0.001, "step": 14310 }, { "epoch": 2.06, "learning_rate": 2.6879156983732672e-05, "loss": 0.0011, "step": 14315 }, { "epoch": 2.07, "learning_rate": 2.6869140155461174e-05, "loss": 0.0008, "step": 14320 }, { "epoch": 2.07, "learning_rate": 2.685912332718968e-05, "loss": 0.001, "step": 14325 }, { "epoch": 2.07, "learning_rate": 2.6849106498918188e-05, "loss": 0.0008, "step": 14330 }, { "epoch": 2.07, "learning_rate": 2.683908967064669e-05, "loss": 0.0009, "step": 14335 }, { "epoch": 2.07, "learning_rate": 2.682907284237519e-05, "loss": 0.0007, "step": 14340 }, { "epoch": 2.07, "learning_rate": 2.6819056014103693e-05, "loss": 0.0007, "step": 14345 }, { "epoch": 2.07, "learning_rate": 2.6809039185832202e-05, "loss": 0.0009, "step": 14350 }, { "epoch": 2.07, "learning_rate": 2.6799022357560704e-05, "loss": 0.0006, "step": 14355 }, { "epoch": 2.07, "learning_rate": 2.678900552928921e-05, "loss": 0.0012, "step": 14360 }, { "epoch": 2.07, "learning_rate": 2.677898870101771e-05, "loss": 0.0012, "step": 14365 }, { "epoch": 2.07, "learning_rate": 2.6768971872746213e-05, "loss": 0.0011, "step": 14370 }, { "epoch": 2.07, "learning_rate": 2.675895504447472e-05, "loss": 0.0006, "step": 14375 }, { "epoch": 2.07, "learning_rate": 2.6748938216203223e-05, "loss": 0.0005, "step": 14380 }, { "epoch": 2.07, "learning_rate": 2.6738921387931725e-05, "loss": 0.0008, "step": 14385 }, { "epoch": 2.08, "learning_rate": 2.6728904559660227e-05, "loss": 0.0007, "step": 14390 }, { "epoch": 2.08, "learning_rate": 2.6718887731388735e-05, "loss": 0.0008, "step": 14395 }, { "epoch": 2.08, "learning_rate": 2.670887090311724e-05, "loss": 0.0012, "step": 14400 }, { "epoch": 2.08, "learning_rate": 2.6698854074845742e-05, "loss": 0.0011, "step": 14405 }, { "epoch": 2.08, "learning_rate": 2.6688837246574244e-05, "loss": 0.0008, "step": 14410 }, { "epoch": 2.08, "learning_rate": 2.6678820418302753e-05, "loss": 0.0006, "step": 14415 }, { "epoch": 2.08, "learning_rate": 2.6668803590031255e-05, "loss": 0.0007, "step": 14420 }, { "epoch": 2.08, "learning_rate": 2.6658786761759757e-05, "loss": 0.0007, "step": 14425 }, { "epoch": 2.08, "learning_rate": 2.6648769933488262e-05, "loss": 0.0008, "step": 14430 }, { "epoch": 2.08, "learning_rate": 2.6638753105216764e-05, "loss": 0.0007, "step": 14435 }, { "epoch": 2.08, "learning_rate": 2.6628736276945272e-05, "loss": 0.0011, "step": 14440 }, { "epoch": 2.08, "learning_rate": 2.6618719448673774e-05, "loss": 0.0007, "step": 14445 }, { "epoch": 2.08, "learning_rate": 2.6608702620402276e-05, "loss": 0.0009, "step": 14450 }, { "epoch": 2.08, "learning_rate": 2.6598685792130778e-05, "loss": 0.0008, "step": 14455 }, { "epoch": 2.09, "learning_rate": 2.6588668963859286e-05, "loss": 0.0006, "step": 14460 }, { "epoch": 2.09, "learning_rate": 2.6578652135587788e-05, "loss": 0.0009, "step": 14465 }, { "epoch": 2.09, "learning_rate": 2.6568635307316293e-05, "loss": 0.0009, "step": 14470 }, { "epoch": 2.09, "learning_rate": 2.6558618479044795e-05, "loss": 0.0007, "step": 14475 }, { "epoch": 2.09, "learning_rate": 2.6548601650773304e-05, "loss": 0.0006, "step": 14480 }, { "epoch": 2.09, "learning_rate": 2.6538584822501806e-05, "loss": 0.001, "step": 14485 }, { "epoch": 2.09, "learning_rate": 2.6528567994230308e-05, "loss": 0.001, "step": 14490 }, { "epoch": 2.09, "learning_rate": 2.651855116595881e-05, "loss": 0.0012, "step": 14495 }, { "epoch": 2.09, "learning_rate": 2.6508534337687318e-05, "loss": 0.0007, "step": 14500 }, { "epoch": 2.09, "learning_rate": 2.649851750941582e-05, "loss": 0.0012, "step": 14505 }, { "epoch": 2.09, "learning_rate": 2.6488500681144325e-05, "loss": 0.0011, "step": 14510 }, { "epoch": 2.09, "learning_rate": 2.6478483852872827e-05, "loss": 0.0006, "step": 14515 }, { "epoch": 2.09, "learning_rate": 2.646846702460133e-05, "loss": 0.0008, "step": 14520 }, { "epoch": 2.1, "learning_rate": 2.6458450196329837e-05, "loss": 0.0011, "step": 14525 }, { "epoch": 2.1, "learning_rate": 2.644843336805834e-05, "loss": 0.001, "step": 14530 }, { "epoch": 2.1, "learning_rate": 2.643841653978684e-05, "loss": 0.001, "step": 14535 }, { "epoch": 2.1, "learning_rate": 2.6428399711515346e-05, "loss": 0.0008, "step": 14540 }, { "epoch": 2.1, "learning_rate": 2.641838288324385e-05, "loss": 0.0008, "step": 14545 }, { "epoch": 2.1, "learning_rate": 2.6408366054972357e-05, "loss": 0.0008, "step": 14550 }, { "epoch": 2.1, "learning_rate": 2.639834922670086e-05, "loss": 0.0009, "step": 14555 }, { "epoch": 2.1, "learning_rate": 2.638833239842936e-05, "loss": 0.001, "step": 14560 }, { "epoch": 2.1, "learning_rate": 2.637831557015787e-05, "loss": 0.0007, "step": 14565 }, { "epoch": 2.1, "learning_rate": 2.636829874188637e-05, "loss": 0.0009, "step": 14570 }, { "epoch": 2.1, "learning_rate": 2.6358281913614873e-05, "loss": 0.001, "step": 14575 }, { "epoch": 2.1, "learning_rate": 2.6348265085343378e-05, "loss": 0.0009, "step": 14580 }, { "epoch": 2.1, "learning_rate": 2.6338248257071883e-05, "loss": 0.0004, "step": 14585 }, { "epoch": 2.1, "learning_rate": 2.632823142880039e-05, "loss": 0.0008, "step": 14590 }, { "epoch": 2.11, "learning_rate": 2.631821460052889e-05, "loss": 0.0005, "step": 14595 }, { "epoch": 2.11, "learning_rate": 2.6308197772257392e-05, "loss": 0.0008, "step": 14600 }, { "epoch": 2.11, "learning_rate": 2.6298180943985894e-05, "loss": 0.0011, "step": 14605 }, { "epoch": 2.11, "learning_rate": 2.6288164115714403e-05, "loss": 0.0008, "step": 14610 }, { "epoch": 2.11, "learning_rate": 2.6278147287442904e-05, "loss": 0.0009, "step": 14615 }, { "epoch": 2.11, "learning_rate": 2.626813045917141e-05, "loss": 0.0009, "step": 14620 }, { "epoch": 2.11, "learning_rate": 2.625811363089991e-05, "loss": 0.0011, "step": 14625 }, { "epoch": 2.11, "learning_rate": 2.624809680262842e-05, "loss": 0.0012, "step": 14630 }, { "epoch": 2.11, "learning_rate": 2.6238079974356922e-05, "loss": 0.0008, "step": 14635 }, { "epoch": 2.11, "learning_rate": 2.6228063146085424e-05, "loss": 0.0006, "step": 14640 }, { "epoch": 2.11, "learning_rate": 2.6218046317813926e-05, "loss": 0.0008, "step": 14645 }, { "epoch": 2.11, "learning_rate": 2.6208029489542434e-05, "loss": 0.001, "step": 14650 }, { "epoch": 2.11, "learning_rate": 2.6198012661270936e-05, "loss": 0.0008, "step": 14655 }, { "epoch": 2.11, "learning_rate": 2.618799583299944e-05, "loss": 0.0003, "step": 14660 }, { "epoch": 2.12, "learning_rate": 2.6177979004727943e-05, "loss": 0.0011, "step": 14665 }, { "epoch": 2.12, "learning_rate": 2.6167962176456445e-05, "loss": 0.0012, "step": 14670 }, { "epoch": 2.12, "learning_rate": 2.6157945348184954e-05, "loss": 0.0006, "step": 14675 }, { "epoch": 2.12, "learning_rate": 2.6147928519913455e-05, "loss": 0.0005, "step": 14680 }, { "epoch": 2.12, "learning_rate": 2.6137911691641957e-05, "loss": 0.0008, "step": 14685 }, { "epoch": 2.12, "learning_rate": 2.6127894863370462e-05, "loss": 0.0009, "step": 14690 }, { "epoch": 2.12, "learning_rate": 2.611787803509897e-05, "loss": 0.0005, "step": 14695 }, { "epoch": 2.12, "learning_rate": 2.6107861206827473e-05, "loss": 0.0009, "step": 14700 }, { "epoch": 2.12, "learning_rate": 2.6097844378555975e-05, "loss": 0.0007, "step": 14705 }, { "epoch": 2.12, "learning_rate": 2.6087827550284477e-05, "loss": 0.0005, "step": 14710 }, { "epoch": 2.12, "learning_rate": 2.6077810722012985e-05, "loss": 0.0007, "step": 14715 }, { "epoch": 2.12, "learning_rate": 2.6067793893741487e-05, "loss": 0.0008, "step": 14720 }, { "epoch": 2.12, "learning_rate": 2.605777706546999e-05, "loss": 0.0005, "step": 14725 }, { "epoch": 2.12, "learning_rate": 2.6047760237198494e-05, "loss": 0.0005, "step": 14730 }, { "epoch": 2.13, "learning_rate": 2.6037743408927003e-05, "loss": 0.0008, "step": 14735 }, { "epoch": 2.13, "learning_rate": 2.6027726580655505e-05, "loss": 0.0006, "step": 14740 }, { "epoch": 2.13, "learning_rate": 2.6017709752384006e-05, "loss": 0.0008, "step": 14745 }, { "epoch": 2.13, "learning_rate": 2.6007692924112508e-05, "loss": 0.001, "step": 14750 }, { "epoch": 2.13, "learning_rate": 2.599767609584101e-05, "loss": 0.0006, "step": 14755 }, { "epoch": 2.13, "learning_rate": 2.598765926756952e-05, "loss": 0.0009, "step": 14760 }, { "epoch": 2.13, "learning_rate": 2.5977642439298024e-05, "loss": 0.0007, "step": 14765 }, { "epoch": 2.13, "learning_rate": 2.5967625611026526e-05, "loss": 0.0008, "step": 14770 }, { "epoch": 2.13, "learning_rate": 2.5957608782755028e-05, "loss": 0.0011, "step": 14775 }, { "epoch": 2.13, "learning_rate": 2.5947591954483536e-05, "loss": 0.001, "step": 14780 }, { "epoch": 2.13, "learning_rate": 2.5937575126212038e-05, "loss": 0.0006, "step": 14785 }, { "epoch": 2.13, "learning_rate": 2.592755829794054e-05, "loss": 0.0008, "step": 14790 }, { "epoch": 2.13, "learning_rate": 2.591754146966904e-05, "loss": 0.0008, "step": 14795 }, { "epoch": 2.13, "learning_rate": 2.590752464139755e-05, "loss": 0.0009, "step": 14800 }, { "epoch": 2.14, "learning_rate": 2.5897507813126056e-05, "loss": 0.0008, "step": 14805 }, { "epoch": 2.14, "learning_rate": 2.5887490984854557e-05, "loss": 0.0006, "step": 14810 }, { "epoch": 2.14, "learning_rate": 2.587747415658306e-05, "loss": 0.0008, "step": 14815 }, { "epoch": 2.14, "learning_rate": 2.5867457328311568e-05, "loss": 0.0005, "step": 14820 }, { "epoch": 2.14, "learning_rate": 2.585744050004007e-05, "loss": 0.001, "step": 14825 }, { "epoch": 2.14, "learning_rate": 2.584742367176857e-05, "loss": 0.001, "step": 14830 }, { "epoch": 2.14, "learning_rate": 2.5837406843497077e-05, "loss": 0.001, "step": 14835 }, { "epoch": 2.14, "learning_rate": 2.582739001522558e-05, "loss": 0.0009, "step": 14840 }, { "epoch": 2.14, "learning_rate": 2.5817373186954087e-05, "loss": 0.0009, "step": 14845 }, { "epoch": 2.14, "learning_rate": 2.580735635868259e-05, "loss": 0.0006, "step": 14850 }, { "epoch": 2.14, "learning_rate": 2.579733953041109e-05, "loss": 0.0009, "step": 14855 }, { "epoch": 2.14, "learning_rate": 2.5787322702139593e-05, "loss": 0.0009, "step": 14860 }, { "epoch": 2.14, "learning_rate": 2.57773058738681e-05, "loss": 0.0006, "step": 14865 }, { "epoch": 2.14, "learning_rate": 2.5767289045596603e-05, "loss": 0.0009, "step": 14870 }, { "epoch": 2.15, "learning_rate": 2.575727221732511e-05, "loss": 0.0005, "step": 14875 }, { "epoch": 2.15, "learning_rate": 2.574725538905361e-05, "loss": 0.0007, "step": 14880 }, { "epoch": 2.15, "learning_rate": 2.573723856078212e-05, "loss": 0.0007, "step": 14885 }, { "epoch": 2.15, "learning_rate": 2.572722173251062e-05, "loss": 0.0008, "step": 14890 }, { "epoch": 2.15, "learning_rate": 2.5717204904239123e-05, "loss": 0.0006, "step": 14895 }, { "epoch": 2.15, "learning_rate": 2.5707188075967624e-05, "loss": 0.0007, "step": 14900 }, { "epoch": 2.15, "learning_rate": 2.5697171247696133e-05, "loss": 0.0009, "step": 14905 }, { "epoch": 2.15, "learning_rate": 2.5687154419424635e-05, "loss": 0.0005, "step": 14910 }, { "epoch": 2.15, "learning_rate": 2.567713759115314e-05, "loss": 0.0011, "step": 14915 }, { "epoch": 2.15, "learning_rate": 2.5667120762881642e-05, "loss": 0.0008, "step": 14920 }, { "epoch": 2.15, "learning_rate": 2.5657103934610144e-05, "loss": 0.0005, "step": 14925 }, { "epoch": 2.15, "learning_rate": 2.5647087106338652e-05, "loss": 0.0008, "step": 14930 }, { "epoch": 2.15, "learning_rate": 2.5637070278067154e-05, "loss": 0.0006, "step": 14935 }, { "epoch": 2.15, "learning_rate": 2.5627053449795656e-05, "loss": 0.0007, "step": 14940 }, { "epoch": 2.16, "learning_rate": 2.561703662152416e-05, "loss": 0.0007, "step": 14945 }, { "epoch": 2.16, "learning_rate": 2.5607019793252666e-05, "loss": 0.0007, "step": 14950 }, { "epoch": 2.16, "learning_rate": 2.559700296498117e-05, "loss": 0.001, "step": 14955 }, { "epoch": 2.16, "learning_rate": 2.5586986136709674e-05, "loss": 0.0008, "step": 14960 }, { "epoch": 2.16, "learning_rate": 2.5576969308438175e-05, "loss": 0.0012, "step": 14965 }, { "epoch": 2.16, "learning_rate": 2.5566952480166684e-05, "loss": 0.001, "step": 14970 }, { "epoch": 2.16, "learning_rate": 2.5556935651895186e-05, "loss": 0.0004, "step": 14975 }, { "epoch": 2.16, "learning_rate": 2.5546918823623688e-05, "loss": 0.001, "step": 14980 }, { "epoch": 2.16, "learning_rate": 2.5536901995352193e-05, "loss": 0.0006, "step": 14985 }, { "epoch": 2.16, "learning_rate": 2.5526885167080695e-05, "loss": 0.0007, "step": 14990 }, { "epoch": 2.16, "learning_rate": 2.5516868338809203e-05, "loss": 0.0011, "step": 14995 }, { "epoch": 2.16, "learning_rate": 2.5506851510537705e-05, "loss": 0.001, "step": 15000 }, { "epoch": 2.16, "learning_rate": 2.5496834682266207e-05, "loss": 0.0007, "step": 15005 }, { "epoch": 2.17, "learning_rate": 2.548681785399471e-05, "loss": 0.0007, "step": 15010 }, { "epoch": 2.17, "learning_rate": 2.5476801025723217e-05, "loss": 0.0007, "step": 15015 }, { "epoch": 2.17, "learning_rate": 2.546678419745172e-05, "loss": 0.0013, "step": 15020 }, { "epoch": 2.17, "learning_rate": 2.5456767369180225e-05, "loss": 0.0006, "step": 15025 }, { "epoch": 2.17, "learning_rate": 2.5446750540908726e-05, "loss": 0.0009, "step": 15030 }, { "epoch": 2.17, "learning_rate": 2.5436733712637235e-05, "loss": 0.0005, "step": 15035 }, { "epoch": 2.17, "learning_rate": 2.5426716884365737e-05, "loss": 0.001, "step": 15040 }, { "epoch": 2.17, "learning_rate": 2.541670005609424e-05, "loss": 0.001, "step": 15045 }, { "epoch": 2.17, "learning_rate": 2.540668322782274e-05, "loss": 0.0005, "step": 15050 }, { "epoch": 2.17, "learning_rate": 2.539666639955125e-05, "loss": 0.001, "step": 15055 }, { "epoch": 2.17, "learning_rate": 2.538664957127975e-05, "loss": 0.0004, "step": 15060 }, { "epoch": 2.17, "learning_rate": 2.5376632743008256e-05, "loss": 0.0008, "step": 15065 }, { "epoch": 2.17, "learning_rate": 2.5366615914736758e-05, "loss": 0.0008, "step": 15070 }, { "epoch": 2.17, "learning_rate": 2.535659908646526e-05, "loss": 0.0005, "step": 15075 }, { "epoch": 2.18, "learning_rate": 2.534658225819377e-05, "loss": 0.001, "step": 15080 }, { "epoch": 2.18, "learning_rate": 2.533656542992227e-05, "loss": 0.0008, "step": 15085 }, { "epoch": 2.18, "learning_rate": 2.5326548601650772e-05, "loss": 0.0009, "step": 15090 }, { "epoch": 2.18, "learning_rate": 2.5316531773379277e-05, "loss": 0.0009, "step": 15095 }, { "epoch": 2.18, "learning_rate": 2.5306514945107786e-05, "loss": 0.001, "step": 15100 }, { "epoch": 2.18, "learning_rate": 2.5296498116836288e-05, "loss": 0.0007, "step": 15105 }, { "epoch": 2.18, "learning_rate": 2.528648128856479e-05, "loss": 0.0012, "step": 15110 }, { "epoch": 2.18, "learning_rate": 2.527646446029329e-05, "loss": 0.0007, "step": 15115 }, { "epoch": 2.18, "learning_rate": 2.52664476320218e-05, "loss": 0.0009, "step": 15120 }, { "epoch": 2.18, "learning_rate": 2.5256430803750302e-05, "loss": 0.0009, "step": 15125 }, { "epoch": 2.18, "learning_rate": 2.5246413975478804e-05, "loss": 0.0007, "step": 15130 }, { "epoch": 2.18, "learning_rate": 2.523639714720731e-05, "loss": 0.0007, "step": 15135 }, { "epoch": 2.18, "learning_rate": 2.5226380318935818e-05, "loss": 0.0009, "step": 15140 }, { "epoch": 2.18, "learning_rate": 2.521636349066432e-05, "loss": 0.0007, "step": 15145 }, { "epoch": 2.19, "learning_rate": 2.520634666239282e-05, "loss": 0.001, "step": 15150 }, { "epoch": 2.19, "learning_rate": 2.5196329834121323e-05, "loss": 0.0009, "step": 15155 }, { "epoch": 2.19, "learning_rate": 2.5186313005849825e-05, "loss": 0.0007, "step": 15160 }, { "epoch": 2.19, "learning_rate": 2.5176296177578334e-05, "loss": 0.0009, "step": 15165 }, { "epoch": 2.19, "learning_rate": 2.516627934930684e-05, "loss": 0.0009, "step": 15170 }, { "epoch": 2.19, "learning_rate": 2.515626252103534e-05, "loss": 0.0009, "step": 15175 }, { "epoch": 2.19, "learning_rate": 2.5146245692763843e-05, "loss": 0.0011, "step": 15180 }, { "epoch": 2.19, "learning_rate": 2.513622886449235e-05, "loss": 0.0011, "step": 15185 }, { "epoch": 2.19, "learning_rate": 2.5126212036220853e-05, "loss": 0.0008, "step": 15190 }, { "epoch": 2.19, "learning_rate": 2.5116195207949355e-05, "loss": 0.0011, "step": 15195 }, { "epoch": 2.19, "learning_rate": 2.5106178379677857e-05, "loss": 0.0006, "step": 15200 }, { "epoch": 2.19, "learning_rate": 2.5096161551406365e-05, "loss": 0.0006, "step": 15205 }, { "epoch": 2.19, "learning_rate": 2.508614472313487e-05, "loss": 0.0006, "step": 15210 }, { "epoch": 2.19, "learning_rate": 2.5076127894863372e-05, "loss": 0.0009, "step": 15215 }, { "epoch": 2.2, "learning_rate": 2.5066111066591874e-05, "loss": 0.0006, "step": 15220 }, { "epoch": 2.2, "learning_rate": 2.5056094238320376e-05, "loss": 0.0008, "step": 15225 }, { "epoch": 2.2, "learning_rate": 2.5046077410048885e-05, "loss": 0.001, "step": 15230 }, { "epoch": 2.2, "learning_rate": 2.5036060581777386e-05, "loss": 0.0006, "step": 15235 }, { "epoch": 2.2, "learning_rate": 2.502604375350589e-05, "loss": 0.0011, "step": 15240 }, { "epoch": 2.2, "learning_rate": 2.5016026925234394e-05, "loss": 0.0006, "step": 15245 }, { "epoch": 2.2, "learning_rate": 2.5006010096962902e-05, "loss": 0.0009, "step": 15250 }, { "epoch": 2.2, "learning_rate": 2.4995993268691404e-05, "loss": 0.0007, "step": 15255 }, { "epoch": 2.2, "learning_rate": 2.4985976440419906e-05, "loss": 0.0006, "step": 15260 }, { "epoch": 2.2, "learning_rate": 2.497595961214841e-05, "loss": 0.0008, "step": 15265 }, { "epoch": 2.2, "learning_rate": 2.4965942783876913e-05, "loss": 0.001, "step": 15270 }, { "epoch": 2.2, "learning_rate": 2.4955925955605418e-05, "loss": 0.0008, "step": 15275 }, { "epoch": 2.2, "learning_rate": 2.4945909127333923e-05, "loss": 0.0017, "step": 15280 }, { "epoch": 2.2, "learning_rate": 2.493589229906243e-05, "loss": 0.0008, "step": 15285 }, { "epoch": 2.21, "learning_rate": 2.492587547079093e-05, "loss": 0.0005, "step": 15290 }, { "epoch": 2.21, "learning_rate": 2.4915858642519436e-05, "loss": 0.0005, "step": 15295 }, { "epoch": 2.21, "learning_rate": 2.4905841814247937e-05, "loss": 0.0009, "step": 15300 }, { "epoch": 2.21, "learning_rate": 2.489582498597644e-05, "loss": 0.0012, "step": 15305 }, { "epoch": 2.21, "learning_rate": 2.4885808157704945e-05, "loss": 0.0011, "step": 15310 }, { "epoch": 2.21, "learning_rate": 2.487579132943345e-05, "loss": 0.0009, "step": 15315 }, { "epoch": 2.21, "learning_rate": 2.4865774501161955e-05, "loss": 0.0008, "step": 15320 }, { "epoch": 2.21, "learning_rate": 2.4855757672890457e-05, "loss": 0.0007, "step": 15325 }, { "epoch": 2.21, "learning_rate": 2.4845740844618962e-05, "loss": 0.0008, "step": 15330 }, { "epoch": 2.21, "learning_rate": 2.4835724016347464e-05, "loss": 0.0007, "step": 15335 }, { "epoch": 2.21, "learning_rate": 2.482570718807597e-05, "loss": 0.0007, "step": 15340 }, { "epoch": 2.21, "learning_rate": 2.481569035980447e-05, "loss": 0.0007, "step": 15345 }, { "epoch": 2.21, "learning_rate": 2.4805673531532976e-05, "loss": 0.0008, "step": 15350 }, { "epoch": 2.21, "learning_rate": 2.479565670326148e-05, "loss": 0.001, "step": 15355 }, { "epoch": 2.22, "learning_rate": 2.4785639874989987e-05, "loss": 0.0009, "step": 15360 }, { "epoch": 2.22, "learning_rate": 2.477562304671849e-05, "loss": 0.0005, "step": 15365 }, { "epoch": 2.22, "learning_rate": 2.4765606218446994e-05, "loss": 0.0009, "step": 15370 }, { "epoch": 2.22, "learning_rate": 2.4755589390175496e-05, "loss": 0.0006, "step": 15375 }, { "epoch": 2.22, "learning_rate": 2.4745572561903997e-05, "loss": 0.0011, "step": 15380 }, { "epoch": 2.22, "learning_rate": 2.4735555733632503e-05, "loss": 0.0008, "step": 15385 }, { "epoch": 2.22, "learning_rate": 2.4725538905361008e-05, "loss": 0.0009, "step": 15390 }, { "epoch": 2.22, "learning_rate": 2.4715522077089513e-05, "loss": 0.0005, "step": 15395 }, { "epoch": 2.22, "learning_rate": 2.4705505248818015e-05, "loss": 0.0009, "step": 15400 }, { "epoch": 2.22, "learning_rate": 2.469548842054652e-05, "loss": 0.0009, "step": 15405 }, { "epoch": 2.22, "learning_rate": 2.4685471592275022e-05, "loss": 0.0007, "step": 15410 }, { "epoch": 2.22, "learning_rate": 2.4675454764003527e-05, "loss": 0.0005, "step": 15415 }, { "epoch": 2.22, "learning_rate": 2.466543793573203e-05, "loss": 0.0011, "step": 15420 }, { "epoch": 2.22, "learning_rate": 2.4655421107460534e-05, "loss": 0.0005, "step": 15425 }, { "epoch": 2.23, "learning_rate": 2.464540427918904e-05, "loss": 0.0006, "step": 15430 }, { "epoch": 2.23, "learning_rate": 2.4635387450917545e-05, "loss": 0.0007, "step": 15435 }, { "epoch": 2.23, "learning_rate": 2.4625370622646047e-05, "loss": 0.0007, "step": 15440 }, { "epoch": 2.23, "learning_rate": 2.4615353794374552e-05, "loss": 0.0009, "step": 15445 }, { "epoch": 2.23, "learning_rate": 2.4605336966103054e-05, "loss": 0.0008, "step": 15450 }, { "epoch": 2.23, "learning_rate": 2.459532013783156e-05, "loss": 0.0008, "step": 15455 }, { "epoch": 2.23, "learning_rate": 2.458530330956006e-05, "loss": 0.0012, "step": 15460 }, { "epoch": 2.23, "learning_rate": 2.4575286481288566e-05, "loss": 0.0006, "step": 15465 }, { "epoch": 2.23, "learning_rate": 2.456526965301707e-05, "loss": 0.0007, "step": 15470 }, { "epoch": 2.23, "learning_rate": 2.4555252824745573e-05, "loss": 0.001, "step": 15475 }, { "epoch": 2.23, "learning_rate": 2.4545235996474078e-05, "loss": 0.0008, "step": 15480 }, { "epoch": 2.23, "learning_rate": 2.453521916820258e-05, "loss": 0.0007, "step": 15485 }, { "epoch": 2.23, "learning_rate": 2.4525202339931085e-05, "loss": 0.0009, "step": 15490 }, { "epoch": 2.23, "learning_rate": 2.4515185511659587e-05, "loss": 0.0005, "step": 15495 }, { "epoch": 2.24, "learning_rate": 2.4505168683388092e-05, "loss": 0.0009, "step": 15500 }, { "epoch": 2.24, "learning_rate": 2.4495151855116598e-05, "loss": 0.0007, "step": 15505 }, { "epoch": 2.24, "learning_rate": 2.4485135026845103e-05, "loss": 0.0006, "step": 15510 }, { "epoch": 2.24, "learning_rate": 2.4475118198573605e-05, "loss": 0.0014, "step": 15515 }, { "epoch": 2.24, "learning_rate": 2.446510137030211e-05, "loss": 0.0011, "step": 15520 }, { "epoch": 2.24, "learning_rate": 2.445508454203061e-05, "loss": 0.0011, "step": 15525 }, { "epoch": 2.24, "learning_rate": 2.4445067713759117e-05, "loss": 0.0009, "step": 15530 }, { "epoch": 2.24, "learning_rate": 2.443505088548762e-05, "loss": 0.0009, "step": 15535 }, { "epoch": 2.24, "learning_rate": 2.4425034057216124e-05, "loss": 0.0006, "step": 15540 }, { "epoch": 2.24, "learning_rate": 2.441501722894463e-05, "loss": 0.0009, "step": 15545 }, { "epoch": 2.24, "learning_rate": 2.440500040067313e-05, "loss": 0.0009, "step": 15550 }, { "epoch": 2.24, "learning_rate": 2.4394983572401636e-05, "loss": 0.0007, "step": 15555 }, { "epoch": 2.24, "learning_rate": 2.4384966744130138e-05, "loss": 0.0007, "step": 15560 }, { "epoch": 2.25, "learning_rate": 2.4374949915858643e-05, "loss": 0.0006, "step": 15565 }, { "epoch": 2.25, "learning_rate": 2.4364933087587145e-05, "loss": 0.001, "step": 15570 }, { "epoch": 2.25, "learning_rate": 2.4354916259315654e-05, "loss": 0.0006, "step": 15575 }, { "epoch": 2.25, "learning_rate": 2.4344899431044156e-05, "loss": 0.0007, "step": 15580 }, { "epoch": 2.25, "learning_rate": 2.433488260277266e-05, "loss": 0.001, "step": 15585 }, { "epoch": 2.25, "learning_rate": 2.4324865774501163e-05, "loss": 0.0008, "step": 15590 }, { "epoch": 2.25, "learning_rate": 2.4314848946229668e-05, "loss": 0.0008, "step": 15595 }, { "epoch": 2.25, "learning_rate": 2.430483211795817e-05, "loss": 0.0006, "step": 15600 }, { "epoch": 2.25, "learning_rate": 2.4294815289686675e-05, "loss": 0.0009, "step": 15605 }, { "epoch": 2.25, "learning_rate": 2.428479846141518e-05, "loss": 0.0007, "step": 15610 }, { "epoch": 2.25, "learning_rate": 2.4274781633143685e-05, "loss": 0.0007, "step": 15615 }, { "epoch": 2.25, "learning_rate": 2.4264764804872187e-05, "loss": 0.0013, "step": 15620 }, { "epoch": 2.25, "learning_rate": 2.425474797660069e-05, "loss": 0.0008, "step": 15625 }, { "epoch": 2.25, "learning_rate": 2.4244731148329194e-05, "loss": 0.0005, "step": 15630 }, { "epoch": 2.26, "learning_rate": 2.4234714320057696e-05, "loss": 0.0007, "step": 15635 }, { "epoch": 2.26, "learning_rate": 2.42246974917862e-05, "loss": 0.0008, "step": 15640 }, { "epoch": 2.26, "learning_rate": 2.4214680663514707e-05, "loss": 0.0007, "step": 15645 }, { "epoch": 2.26, "learning_rate": 2.4204663835243212e-05, "loss": 0.0009, "step": 15650 }, { "epoch": 2.26, "learning_rate": 2.4194647006971714e-05, "loss": 0.001, "step": 15655 }, { "epoch": 2.26, "learning_rate": 2.418463017870022e-05, "loss": 0.0009, "step": 15660 }, { "epoch": 2.26, "learning_rate": 2.417461335042872e-05, "loss": 0.001, "step": 15665 }, { "epoch": 2.26, "learning_rate": 2.4164596522157226e-05, "loss": 0.0009, "step": 15670 }, { "epoch": 2.26, "learning_rate": 2.4154579693885728e-05, "loss": 0.0011, "step": 15675 }, { "epoch": 2.26, "learning_rate": 2.4144562865614233e-05, "loss": 0.0009, "step": 15680 }, { "epoch": 2.26, "learning_rate": 2.4134546037342738e-05, "loss": 0.0008, "step": 15685 }, { "epoch": 2.26, "learning_rate": 2.4124529209071243e-05, "loss": 0.0008, "step": 15690 }, { "epoch": 2.26, "learning_rate": 2.4114512380799745e-05, "loss": 0.0007, "step": 15695 }, { "epoch": 2.26, "learning_rate": 2.4104495552528247e-05, "loss": 0.0007, "step": 15700 }, { "epoch": 2.27, "learning_rate": 2.4094478724256752e-05, "loss": 0.0007, "step": 15705 }, { "epoch": 2.27, "learning_rate": 2.4084461895985254e-05, "loss": 0.0008, "step": 15710 }, { "epoch": 2.27, "learning_rate": 2.407444506771376e-05, "loss": 0.0007, "step": 15715 }, { "epoch": 2.27, "learning_rate": 2.4064428239442265e-05, "loss": 0.0011, "step": 15720 }, { "epoch": 2.27, "learning_rate": 2.405441141117077e-05, "loss": 0.0011, "step": 15725 }, { "epoch": 2.27, "learning_rate": 2.4044394582899272e-05, "loss": 0.0014, "step": 15730 }, { "epoch": 2.27, "learning_rate": 2.4034377754627777e-05, "loss": 0.0008, "step": 15735 }, { "epoch": 2.27, "learning_rate": 2.402436092635628e-05, "loss": 0.0007, "step": 15740 }, { "epoch": 2.27, "learning_rate": 2.4014344098084784e-05, "loss": 0.0009, "step": 15745 }, { "epoch": 2.27, "learning_rate": 2.4004327269813286e-05, "loss": 0.0007, "step": 15750 }, { "epoch": 2.27, "learning_rate": 2.399431044154179e-05, "loss": 0.0011, "step": 15755 }, { "epoch": 2.27, "learning_rate": 2.3984293613270296e-05, "loss": 0.0007, "step": 15760 }, { "epoch": 2.27, "learning_rate": 2.39742767849988e-05, "loss": 0.0009, "step": 15765 }, { "epoch": 2.27, "learning_rate": 2.3964259956727303e-05, "loss": 0.0008, "step": 15770 }, { "epoch": 2.28, "learning_rate": 2.3954243128455805e-05, "loss": 0.0007, "step": 15775 }, { "epoch": 2.28, "learning_rate": 2.394422630018431e-05, "loss": 0.0009, "step": 15780 }, { "epoch": 2.28, "learning_rate": 2.3934209471912812e-05, "loss": 0.0008, "step": 15785 }, { "epoch": 2.28, "learning_rate": 2.3924192643641318e-05, "loss": 0.0007, "step": 15790 }, { "epoch": 2.28, "learning_rate": 2.3914175815369823e-05, "loss": 0.001, "step": 15795 }, { "epoch": 2.28, "learning_rate": 2.3904158987098328e-05, "loss": 0.0009, "step": 15800 }, { "epoch": 2.28, "learning_rate": 2.389414215882683e-05, "loss": 0.0008, "step": 15805 }, { "epoch": 2.28, "learning_rate": 2.3884125330555335e-05, "loss": 0.0008, "step": 15810 }, { "epoch": 2.28, "learning_rate": 2.3874108502283837e-05, "loss": 0.0004, "step": 15815 }, { "epoch": 2.28, "learning_rate": 2.3864091674012342e-05, "loss": 0.0008, "step": 15820 }, { "epoch": 2.28, "learning_rate": 2.3854074845740844e-05, "loss": 0.0008, "step": 15825 }, { "epoch": 2.28, "learning_rate": 2.384405801746935e-05, "loss": 0.0011, "step": 15830 }, { "epoch": 2.28, "learning_rate": 2.3834041189197854e-05, "loss": 0.0011, "step": 15835 }, { "epoch": 2.28, "learning_rate": 2.382402436092636e-05, "loss": 0.0012, "step": 15840 }, { "epoch": 2.29, "learning_rate": 2.381400753265486e-05, "loss": 0.0009, "step": 15845 }, { "epoch": 2.29, "learning_rate": 2.3803990704383367e-05, "loss": 0.0008, "step": 15850 }, { "epoch": 2.29, "learning_rate": 2.379397387611187e-05, "loss": 0.0009, "step": 15855 }, { "epoch": 2.29, "learning_rate": 2.378395704784037e-05, "loss": 0.0009, "step": 15860 }, { "epoch": 2.29, "learning_rate": 2.3773940219568876e-05, "loss": 0.0011, "step": 15865 }, { "epoch": 2.29, "learning_rate": 2.376392339129738e-05, "loss": 0.0007, "step": 15870 }, { "epoch": 2.29, "learning_rate": 2.3753906563025886e-05, "loss": 0.0007, "step": 15875 }, { "epoch": 2.29, "learning_rate": 2.3743889734754388e-05, "loss": 0.0005, "step": 15880 }, { "epoch": 2.29, "learning_rate": 2.3733872906482893e-05, "loss": 0.0008, "step": 15885 }, { "epoch": 2.29, "learning_rate": 2.3723856078211395e-05, "loss": 0.0007, "step": 15890 }, { "epoch": 2.29, "learning_rate": 2.37138392499399e-05, "loss": 0.0006, "step": 15895 }, { "epoch": 2.29, "learning_rate": 2.3703822421668402e-05, "loss": 0.0011, "step": 15900 }, { "epoch": 2.29, "learning_rate": 2.3693805593396907e-05, "loss": 0.0009, "step": 15905 }, { "epoch": 2.29, "learning_rate": 2.3683788765125412e-05, "loss": 0.0008, "step": 15910 }, { "epoch": 2.3, "learning_rate": 2.3673771936853918e-05, "loss": 0.0007, "step": 15915 }, { "epoch": 2.3, "learning_rate": 2.366375510858242e-05, "loss": 0.001, "step": 15920 }, { "epoch": 2.3, "learning_rate": 2.3653738280310925e-05, "loss": 0.0005, "step": 15925 }, { "epoch": 2.3, "learning_rate": 2.3643721452039427e-05, "loss": 0.0011, "step": 15930 }, { "epoch": 2.3, "learning_rate": 2.363370462376793e-05, "loss": 0.0007, "step": 15935 }, { "epoch": 2.3, "learning_rate": 2.3623687795496434e-05, "loss": 0.0013, "step": 15940 }, { "epoch": 2.3, "learning_rate": 2.361367096722494e-05, "loss": 0.0007, "step": 15945 }, { "epoch": 2.3, "learning_rate": 2.3603654138953444e-05, "loss": 0.0006, "step": 15950 }, { "epoch": 2.3, "learning_rate": 2.3593637310681946e-05, "loss": 0.0008, "step": 15955 }, { "epoch": 2.3, "learning_rate": 2.358362048241045e-05, "loss": 0.0006, "step": 15960 }, { "epoch": 2.3, "learning_rate": 2.3573603654138953e-05, "loss": 0.001, "step": 15965 }, { "epoch": 2.3, "learning_rate": 2.3563586825867458e-05, "loss": 0.0005, "step": 15970 }, { "epoch": 2.3, "learning_rate": 2.355356999759596e-05, "loss": 0.001, "step": 15975 }, { "epoch": 2.3, "learning_rate": 2.354355316932447e-05, "loss": 0.0007, "step": 15980 }, { "epoch": 2.31, "learning_rate": 2.353353634105297e-05, "loss": 0.001, "step": 15985 }, { "epoch": 2.31, "learning_rate": 2.3523519512781476e-05, "loss": 0.001, "step": 15990 }, { "epoch": 2.31, "learning_rate": 2.3513502684509978e-05, "loss": 0.0008, "step": 15995 }, { "epoch": 2.31, "learning_rate": 2.3503485856238483e-05, "loss": 0.0007, "step": 16000 }, { "epoch": 2.31, "learning_rate": 2.3493469027966985e-05, "loss": 0.0008, "step": 16005 }, { "epoch": 2.31, "learning_rate": 2.348345219969549e-05, "loss": 0.0008, "step": 16010 }, { "epoch": 2.31, "learning_rate": 2.3473435371423995e-05, "loss": 0.0004, "step": 16015 }, { "epoch": 2.31, "learning_rate": 2.3463418543152497e-05, "loss": 0.0011, "step": 16020 }, { "epoch": 2.31, "learning_rate": 2.3453401714881002e-05, "loss": 0.0013, "step": 16025 }, { "epoch": 2.31, "learning_rate": 2.3443384886609504e-05, "loss": 0.0009, "step": 16030 }, { "epoch": 2.31, "learning_rate": 2.343336805833801e-05, "loss": 0.0009, "step": 16035 }, { "epoch": 2.31, "learning_rate": 2.342335123006651e-05, "loss": 0.0008, "step": 16040 }, { "epoch": 2.31, "learning_rate": 2.3413334401795016e-05, "loss": 0.0006, "step": 16045 }, { "epoch": 2.32, "learning_rate": 2.340331757352352e-05, "loss": 0.0009, "step": 16050 }, { "epoch": 2.32, "learning_rate": 2.3393300745252027e-05, "loss": 0.0007, "step": 16055 }, { "epoch": 2.32, "learning_rate": 2.338328391698053e-05, "loss": 0.0009, "step": 16060 }, { "epoch": 2.32, "learning_rate": 2.3373267088709034e-05, "loss": 0.0011, "step": 16065 }, { "epoch": 2.32, "learning_rate": 2.3363250260437536e-05, "loss": 0.0007, "step": 16070 }, { "epoch": 2.32, "learning_rate": 2.335323343216604e-05, "loss": 0.0008, "step": 16075 }, { "epoch": 2.32, "learning_rate": 2.3343216603894543e-05, "loss": 0.001, "step": 16080 }, { "epoch": 2.32, "learning_rate": 2.3333199775623048e-05, "loss": 0.0012, "step": 16085 }, { "epoch": 2.32, "learning_rate": 2.3323182947351553e-05, "loss": 0.001, "step": 16090 }, { "epoch": 2.32, "learning_rate": 2.3313166119080055e-05, "loss": 0.0009, "step": 16095 }, { "epoch": 2.32, "learning_rate": 2.330314929080856e-05, "loss": 0.0011, "step": 16100 }, { "epoch": 2.32, "learning_rate": 2.3293132462537062e-05, "loss": 0.0008, "step": 16105 }, { "epoch": 2.32, "learning_rate": 2.3283115634265567e-05, "loss": 0.0008, "step": 16110 }, { "epoch": 2.32, "learning_rate": 2.327309880599407e-05, "loss": 0.0008, "step": 16115 }, { "epoch": 2.33, "learning_rate": 2.3263081977722574e-05, "loss": 0.0006, "step": 16120 }, { "epoch": 2.33, "learning_rate": 2.325306514945108e-05, "loss": 0.0009, "step": 16125 }, { "epoch": 2.33, "learning_rate": 2.3243048321179585e-05, "loss": 0.0008, "step": 16130 }, { "epoch": 2.33, "learning_rate": 2.3233031492908087e-05, "loss": 0.0009, "step": 16135 }, { "epoch": 2.33, "learning_rate": 2.3223014664636592e-05, "loss": 0.0006, "step": 16140 }, { "epoch": 2.33, "learning_rate": 2.3212997836365094e-05, "loss": 0.0009, "step": 16145 }, { "epoch": 2.33, "learning_rate": 2.32029810080936e-05, "loss": 0.0005, "step": 16150 }, { "epoch": 2.33, "learning_rate": 2.31929641798221e-05, "loss": 0.0006, "step": 16155 }, { "epoch": 2.33, "learning_rate": 2.3182947351550606e-05, "loss": 0.0008, "step": 16160 }, { "epoch": 2.33, "learning_rate": 2.317293052327911e-05, "loss": 0.0005, "step": 16165 }, { "epoch": 2.33, "learning_rate": 2.3162913695007613e-05, "loss": 0.001, "step": 16170 }, { "epoch": 2.33, "learning_rate": 2.315289686673612e-05, "loss": 0.001, "step": 16175 }, { "epoch": 2.33, "learning_rate": 2.314288003846462e-05, "loss": 0.0013, "step": 16180 }, { "epoch": 2.33, "learning_rate": 2.3132863210193125e-05, "loss": 0.0008, "step": 16185 }, { "epoch": 2.34, "learning_rate": 2.3122846381921627e-05, "loss": 0.0014, "step": 16190 }, { "epoch": 2.34, "learning_rate": 2.3112829553650132e-05, "loss": 0.0009, "step": 16195 }, { "epoch": 2.34, "learning_rate": 2.3102812725378638e-05, "loss": 0.0006, "step": 16200 }, { "epoch": 2.34, "learning_rate": 2.3092795897107143e-05, "loss": 0.0012, "step": 16205 }, { "epoch": 2.34, "learning_rate": 2.3082779068835645e-05, "loss": 0.0004, "step": 16210 }, { "epoch": 2.34, "learning_rate": 2.307276224056415e-05, "loss": 0.0009, "step": 16215 }, { "epoch": 2.34, "learning_rate": 2.3062745412292652e-05, "loss": 0.001, "step": 16220 }, { "epoch": 2.34, "learning_rate": 2.3052728584021157e-05, "loss": 0.0009, "step": 16225 }, { "epoch": 2.34, "learning_rate": 2.304271175574966e-05, "loss": 0.0008, "step": 16230 }, { "epoch": 2.34, "learning_rate": 2.3032694927478164e-05, "loss": 0.0012, "step": 16235 }, { "epoch": 2.34, "learning_rate": 2.302267809920667e-05, "loss": 0.0011, "step": 16240 }, { "epoch": 2.34, "learning_rate": 2.3012661270935175e-05, "loss": 0.0011, "step": 16245 }, { "epoch": 2.34, "learning_rate": 2.3002644442663676e-05, "loss": 0.0008, "step": 16250 }, { "epoch": 2.34, "learning_rate": 2.2992627614392178e-05, "loss": 0.0009, "step": 16255 }, { "epoch": 2.35, "learning_rate": 2.2982610786120683e-05, "loss": 0.001, "step": 16260 }, { "epoch": 2.35, "learning_rate": 2.2972593957849185e-05, "loss": 0.0007, "step": 16265 }, { "epoch": 2.35, "learning_rate": 2.296257712957769e-05, "loss": 0.0009, "step": 16270 }, { "epoch": 2.35, "learning_rate": 2.2952560301306196e-05, "loss": 0.0007, "step": 16275 }, { "epoch": 2.35, "learning_rate": 2.29425434730347e-05, "loss": 0.0006, "step": 16280 }, { "epoch": 2.35, "learning_rate": 2.2932526644763203e-05, "loss": 0.0009, "step": 16285 }, { "epoch": 2.35, "learning_rate": 2.2922509816491708e-05, "loss": 0.0008, "step": 16290 }, { "epoch": 2.35, "learning_rate": 2.291249298822021e-05, "loss": 0.0009, "step": 16295 }, { "epoch": 2.35, "learning_rate": 2.2902476159948715e-05, "loss": 0.0011, "step": 16300 }, { "epoch": 2.35, "learning_rate": 2.2892459331677217e-05, "loss": 0.0007, "step": 16305 }, { "epoch": 2.35, "learning_rate": 2.2882442503405722e-05, "loss": 0.0008, "step": 16310 }, { "epoch": 2.35, "learning_rate": 2.2872425675134227e-05, "loss": 0.001, "step": 16315 }, { "epoch": 2.35, "learning_rate": 2.2862408846862733e-05, "loss": 0.0009, "step": 16320 }, { "epoch": 2.35, "learning_rate": 2.2852392018591234e-05, "loss": 0.0006, "step": 16325 }, { "epoch": 2.36, "learning_rate": 2.2842375190319736e-05, "loss": 0.0007, "step": 16330 }, { "epoch": 2.36, "learning_rate": 2.283235836204824e-05, "loss": 0.0009, "step": 16335 }, { "epoch": 2.36, "learning_rate": 2.2822341533776743e-05, "loss": 0.001, "step": 16340 }, { "epoch": 2.36, "learning_rate": 2.281232470550525e-05, "loss": 0.0007, "step": 16345 }, { "epoch": 2.36, "learning_rate": 2.2802307877233754e-05, "loss": 0.0008, "step": 16350 }, { "epoch": 2.36, "learning_rate": 2.279229104896226e-05, "loss": 0.0007, "step": 16355 }, { "epoch": 2.36, "learning_rate": 2.278227422069076e-05, "loss": 0.0006, "step": 16360 }, { "epoch": 2.36, "learning_rate": 2.2772257392419266e-05, "loss": 0.0007, "step": 16365 }, { "epoch": 2.36, "learning_rate": 2.2762240564147768e-05, "loss": 0.0008, "step": 16370 }, { "epoch": 2.36, "learning_rate": 2.2752223735876273e-05, "loss": 0.0006, "step": 16375 }, { "epoch": 2.36, "learning_rate": 2.2742206907604775e-05, "loss": 0.0007, "step": 16380 }, { "epoch": 2.36, "learning_rate": 2.2732190079333284e-05, "loss": 0.0009, "step": 16385 }, { "epoch": 2.36, "learning_rate": 2.2722173251061786e-05, "loss": 0.0006, "step": 16390 }, { "epoch": 2.36, "learning_rate": 2.271215642279029e-05, "loss": 0.0009, "step": 16395 }, { "epoch": 2.37, "learning_rate": 2.2702139594518793e-05, "loss": 0.0009, "step": 16400 }, { "epoch": 2.37, "learning_rate": 2.2692122766247294e-05, "loss": 0.0009, "step": 16405 }, { "epoch": 2.37, "learning_rate": 2.26821059379758e-05, "loss": 0.0011, "step": 16410 }, { "epoch": 2.37, "learning_rate": 2.26720891097043e-05, "loss": 0.0005, "step": 16415 }, { "epoch": 2.37, "learning_rate": 2.266207228143281e-05, "loss": 0.0005, "step": 16420 }, { "epoch": 2.37, "learning_rate": 2.2652055453161312e-05, "loss": 0.0009, "step": 16425 }, { "epoch": 2.37, "learning_rate": 2.2642038624889817e-05, "loss": 0.0006, "step": 16430 }, { "epoch": 2.37, "learning_rate": 2.263202179661832e-05, "loss": 0.0012, "step": 16435 }, { "epoch": 2.37, "learning_rate": 2.2622004968346824e-05, "loss": 0.0007, "step": 16440 }, { "epoch": 2.37, "learning_rate": 2.2611988140075326e-05, "loss": 0.0008, "step": 16445 }, { "epoch": 2.37, "learning_rate": 2.260197131180383e-05, "loss": 0.001, "step": 16450 }, { "epoch": 2.37, "learning_rate": 2.2591954483532337e-05, "loss": 0.0008, "step": 16455 }, { "epoch": 2.37, "learning_rate": 2.2581937655260842e-05, "loss": 0.001, "step": 16460 }, { "epoch": 2.37, "learning_rate": 2.2571920826989344e-05, "loss": 0.0008, "step": 16465 }, { "epoch": 2.38, "learning_rate": 2.256190399871785e-05, "loss": 0.0008, "step": 16470 }, { "epoch": 2.38, "learning_rate": 2.255188717044635e-05, "loss": 0.0006, "step": 16475 }, { "epoch": 2.38, "learning_rate": 2.2541870342174856e-05, "loss": 0.0009, "step": 16480 }, { "epoch": 2.38, "learning_rate": 2.2531853513903358e-05, "loss": 0.0008, "step": 16485 }, { "epoch": 2.38, "learning_rate": 2.2521836685631863e-05, "loss": 0.0006, "step": 16490 }, { "epoch": 2.38, "learning_rate": 2.2511819857360368e-05, "loss": 0.0005, "step": 16495 }, { "epoch": 2.38, "learning_rate": 2.250180302908887e-05, "loss": 0.0006, "step": 16500 }, { "epoch": 2.38, "learning_rate": 2.2491786200817375e-05, "loss": 0.001, "step": 16505 }, { "epoch": 2.38, "learning_rate": 2.2481769372545877e-05, "loss": 0.001, "step": 16510 }, { "epoch": 2.38, "learning_rate": 2.2471752544274382e-05, "loss": 0.0008, "step": 16515 }, { "epoch": 2.38, "learning_rate": 2.2461735716002884e-05, "loss": 0.0008, "step": 16520 }, { "epoch": 2.38, "learning_rate": 2.245171888773139e-05, "loss": 0.0009, "step": 16525 }, { "epoch": 2.38, "learning_rate": 2.2441702059459895e-05, "loss": 0.0007, "step": 16530 }, { "epoch": 2.38, "learning_rate": 2.24316852311884e-05, "loss": 0.001, "step": 16535 }, { "epoch": 2.39, "learning_rate": 2.24216684029169e-05, "loss": 0.0007, "step": 16540 }, { "epoch": 2.39, "learning_rate": 2.2411651574645407e-05, "loss": 0.0006, "step": 16545 }, { "epoch": 2.39, "learning_rate": 2.240163474637391e-05, "loss": 0.0011, "step": 16550 }, { "epoch": 2.39, "learning_rate": 2.2391617918102414e-05, "loss": 0.001, "step": 16555 }, { "epoch": 2.39, "learning_rate": 2.2381601089830916e-05, "loss": 0.0009, "step": 16560 }, { "epoch": 2.39, "learning_rate": 2.237158426155942e-05, "loss": 0.0007, "step": 16565 }, { "epoch": 2.39, "learning_rate": 2.2361567433287926e-05, "loss": 0.0009, "step": 16570 }, { "epoch": 2.39, "learning_rate": 2.2351550605016428e-05, "loss": 0.001, "step": 16575 }, { "epoch": 2.39, "learning_rate": 2.2341533776744933e-05, "loss": 0.0008, "step": 16580 }, { "epoch": 2.39, "learning_rate": 2.2331516948473435e-05, "loss": 0.0009, "step": 16585 }, { "epoch": 2.39, "learning_rate": 2.232150012020194e-05, "loss": 0.0009, "step": 16590 }, { "epoch": 2.39, "learning_rate": 2.2311483291930442e-05, "loss": 0.0006, "step": 16595 }, { "epoch": 2.39, "learning_rate": 2.2301466463658947e-05, "loss": 0.0009, "step": 16600 }, { "epoch": 2.4, "learning_rate": 2.2291449635387453e-05, "loss": 0.0006, "step": 16605 }, { "epoch": 2.4, "learning_rate": 2.2281432807115958e-05, "loss": 0.0008, "step": 16610 }, { "epoch": 2.4, "learning_rate": 2.227141597884446e-05, "loss": 0.001, "step": 16615 }, { "epoch": 2.4, "learning_rate": 2.2261399150572965e-05, "loss": 0.0006, "step": 16620 }, { "epoch": 2.4, "learning_rate": 2.2251382322301467e-05, "loss": 0.001, "step": 16625 }, { "epoch": 2.4, "learning_rate": 2.2241365494029972e-05, "loss": 0.0005, "step": 16630 }, { "epoch": 2.4, "learning_rate": 2.2231348665758474e-05, "loss": 0.0009, "step": 16635 }, { "epoch": 2.4, "learning_rate": 2.222133183748698e-05, "loss": 0.0011, "step": 16640 }, { "epoch": 2.4, "learning_rate": 2.2211315009215484e-05, "loss": 0.0011, "step": 16645 }, { "epoch": 2.4, "learning_rate": 2.2201298180943986e-05, "loss": 0.0012, "step": 16650 }, { "epoch": 2.4, "learning_rate": 2.219128135267249e-05, "loss": 0.0007, "step": 16655 }, { "epoch": 2.4, "learning_rate": 2.2181264524400993e-05, "loss": 0.0012, "step": 16660 }, { "epoch": 2.4, "learning_rate": 2.21712476961295e-05, "loss": 0.0008, "step": 16665 }, { "epoch": 2.4, "learning_rate": 2.2161230867858e-05, "loss": 0.0007, "step": 16670 }, { "epoch": 2.41, "learning_rate": 2.2151214039586506e-05, "loss": 0.0007, "step": 16675 }, { "epoch": 2.41, "learning_rate": 2.214119721131501e-05, "loss": 0.0009, "step": 16680 }, { "epoch": 2.41, "learning_rate": 2.2131180383043516e-05, "loss": 0.0007, "step": 16685 }, { "epoch": 2.41, "learning_rate": 2.2121163554772018e-05, "loss": 0.0006, "step": 16690 }, { "epoch": 2.41, "learning_rate": 2.2111146726500523e-05, "loss": 0.0008, "step": 16695 }, { "epoch": 2.41, "learning_rate": 2.2101129898229025e-05, "loss": 0.0005, "step": 16700 }, { "epoch": 2.41, "learning_rate": 2.209111306995753e-05, "loss": 0.0006, "step": 16705 }, { "epoch": 2.41, "learning_rate": 2.2081096241686032e-05, "loss": 0.0007, "step": 16710 }, { "epoch": 2.41, "learning_rate": 2.2071079413414537e-05, "loss": 0.0011, "step": 16715 }, { "epoch": 2.41, "learning_rate": 2.2061062585143042e-05, "loss": 0.0007, "step": 16720 }, { "epoch": 2.41, "learning_rate": 2.2051045756871544e-05, "loss": 0.0008, "step": 16725 }, { "epoch": 2.41, "learning_rate": 2.204102892860005e-05, "loss": 0.0009, "step": 16730 }, { "epoch": 2.41, "learning_rate": 2.203101210032855e-05, "loss": 0.0009, "step": 16735 }, { "epoch": 2.41, "learning_rate": 2.2020995272057057e-05, "loss": 0.0007, "step": 16740 }, { "epoch": 2.42, "learning_rate": 2.201097844378556e-05, "loss": 0.0005, "step": 16745 }, { "epoch": 2.42, "learning_rate": 2.2000961615514064e-05, "loss": 0.0006, "step": 16750 }, { "epoch": 2.42, "learning_rate": 2.199094478724257e-05, "loss": 0.0006, "step": 16755 }, { "epoch": 2.42, "learning_rate": 2.1980927958971074e-05, "loss": 0.0008, "step": 16760 }, { "epoch": 2.42, "learning_rate": 2.1970911130699576e-05, "loss": 0.001, "step": 16765 }, { "epoch": 2.42, "learning_rate": 2.196089430242808e-05, "loss": 0.0008, "step": 16770 }, { "epoch": 2.42, "learning_rate": 2.1950877474156583e-05, "loss": 0.0008, "step": 16775 }, { "epoch": 2.42, "learning_rate": 2.1940860645885088e-05, "loss": 0.0007, "step": 16780 }, { "epoch": 2.42, "learning_rate": 2.193084381761359e-05, "loss": 0.0006, "step": 16785 }, { "epoch": 2.42, "learning_rate": 2.19208269893421e-05, "loss": 0.0012, "step": 16790 }, { "epoch": 2.42, "learning_rate": 2.19108101610706e-05, "loss": 0.0007, "step": 16795 }, { "epoch": 2.42, "learning_rate": 2.1900793332799102e-05, "loss": 0.0007, "step": 16800 }, { "epoch": 2.42, "learning_rate": 2.1890776504527608e-05, "loss": 0.0005, "step": 16805 }, { "epoch": 2.42, "learning_rate": 2.188075967625611e-05, "loss": 0.0009, "step": 16810 }, { "epoch": 2.43, "learning_rate": 2.1870742847984615e-05, "loss": 0.001, "step": 16815 }, { "epoch": 2.43, "learning_rate": 2.1860726019713116e-05, "loss": 0.0005, "step": 16820 }, { "epoch": 2.43, "learning_rate": 2.1850709191441625e-05, "loss": 0.0009, "step": 16825 }, { "epoch": 2.43, "learning_rate": 2.1840692363170127e-05, "loss": 0.0005, "step": 16830 }, { "epoch": 2.43, "learning_rate": 2.1830675534898632e-05, "loss": 0.0007, "step": 16835 }, { "epoch": 2.43, "learning_rate": 2.1820658706627134e-05, "loss": 0.0007, "step": 16840 }, { "epoch": 2.43, "learning_rate": 2.181064187835564e-05, "loss": 0.0006, "step": 16845 }, { "epoch": 2.43, "learning_rate": 2.180062505008414e-05, "loss": 0.0008, "step": 16850 }, { "epoch": 2.43, "learning_rate": 2.1790608221812646e-05, "loss": 0.0009, "step": 16855 }, { "epoch": 2.43, "learning_rate": 2.178059139354115e-05, "loss": 0.0008, "step": 16860 }, { "epoch": 2.43, "learning_rate": 2.1770574565269657e-05, "loss": 0.0008, "step": 16865 }, { "epoch": 2.43, "learning_rate": 2.176055773699816e-05, "loss": 0.0011, "step": 16870 }, { "epoch": 2.43, "learning_rate": 2.1750540908726664e-05, "loss": 0.0012, "step": 16875 }, { "epoch": 2.43, "learning_rate": 2.1740524080455166e-05, "loss": 0.0008, "step": 16880 }, { "epoch": 2.44, "learning_rate": 2.1730507252183667e-05, "loss": 0.0007, "step": 16885 }, { "epoch": 2.44, "learning_rate": 2.1720490423912173e-05, "loss": 0.0006, "step": 16890 }, { "epoch": 2.44, "learning_rate": 2.1710473595640678e-05, "loss": 0.0011, "step": 16895 }, { "epoch": 2.44, "learning_rate": 2.1700456767369183e-05, "loss": 0.0009, "step": 16900 }, { "epoch": 2.44, "learning_rate": 2.1690439939097685e-05, "loss": 0.0007, "step": 16905 }, { "epoch": 2.44, "learning_rate": 2.168042311082619e-05, "loss": 0.0008, "step": 16910 }, { "epoch": 2.44, "learning_rate": 2.1670406282554692e-05, "loss": 0.0011, "step": 16915 }, { "epoch": 2.44, "learning_rate": 2.1660389454283197e-05, "loss": 0.0009, "step": 16920 }, { "epoch": 2.44, "learning_rate": 2.16503726260117e-05, "loss": 0.0006, "step": 16925 }, { "epoch": 2.44, "learning_rate": 2.1640355797740204e-05, "loss": 0.0009, "step": 16930 }, { "epoch": 2.44, "learning_rate": 2.163033896946871e-05, "loss": 0.001, "step": 16935 }, { "epoch": 2.44, "learning_rate": 2.1620322141197215e-05, "loss": 0.0005, "step": 16940 }, { "epoch": 2.44, "learning_rate": 2.1610305312925717e-05, "loss": 0.0009, "step": 16945 }, { "epoch": 2.44, "learning_rate": 2.1600288484654222e-05, "loss": 0.0008, "step": 16950 }, { "epoch": 2.45, "learning_rate": 2.1590271656382724e-05, "loss": 0.0009, "step": 16955 }, { "epoch": 2.45, "learning_rate": 2.1580254828111226e-05, "loss": 0.0007, "step": 16960 }, { "epoch": 2.45, "learning_rate": 2.157023799983973e-05, "loss": 0.0005, "step": 16965 }, { "epoch": 2.45, "learning_rate": 2.1560221171568236e-05, "loss": 0.0006, "step": 16970 }, { "epoch": 2.45, "learning_rate": 2.155020434329674e-05, "loss": 0.0009, "step": 16975 }, { "epoch": 2.45, "learning_rate": 2.1540187515025243e-05, "loss": 0.0008, "step": 16980 }, { "epoch": 2.45, "learning_rate": 2.1530170686753748e-05, "loss": 0.0008, "step": 16985 }, { "epoch": 2.45, "learning_rate": 2.152015385848225e-05, "loss": 0.0011, "step": 16990 }, { "epoch": 2.45, "learning_rate": 2.1510137030210755e-05, "loss": 0.001, "step": 16995 }, { "epoch": 2.45, "learning_rate": 2.1500120201939257e-05, "loss": 0.001, "step": 17000 }, { "epoch": 2.45, "learning_rate": 2.1490103373667762e-05, "loss": 0.0011, "step": 17005 }, { "epoch": 2.45, "learning_rate": 2.1480086545396268e-05, "loss": 0.0009, "step": 17010 }, { "epoch": 2.45, "learning_rate": 2.1470069717124773e-05, "loss": 0.0009, "step": 17015 }, { "epoch": 2.45, "learning_rate": 2.1460052888853275e-05, "loss": 0.0007, "step": 17020 }, { "epoch": 2.46, "learning_rate": 2.145003606058178e-05, "loss": 0.0012, "step": 17025 }, { "epoch": 2.46, "learning_rate": 2.1440019232310282e-05, "loss": 0.0007, "step": 17030 }, { "epoch": 2.46, "learning_rate": 2.1430002404038787e-05, "loss": 0.001, "step": 17035 }, { "epoch": 2.46, "learning_rate": 2.141998557576729e-05, "loss": 0.0009, "step": 17040 }, { "epoch": 2.46, "learning_rate": 2.1409968747495794e-05, "loss": 0.0009, "step": 17045 }, { "epoch": 2.46, "learning_rate": 2.13999519192243e-05, "loss": 0.0006, "step": 17050 }, { "epoch": 2.46, "learning_rate": 2.13899350909528e-05, "loss": 0.001, "step": 17055 }, { "epoch": 2.46, "learning_rate": 2.1379918262681306e-05, "loss": 0.0007, "step": 17060 }, { "epoch": 2.46, "learning_rate": 2.1369901434409808e-05, "loss": 0.0012, "step": 17065 }, { "epoch": 2.46, "learning_rate": 2.1359884606138313e-05, "loss": 0.0008, "step": 17070 }, { "epoch": 2.46, "learning_rate": 2.1349867777866815e-05, "loss": 0.001, "step": 17075 }, { "epoch": 2.46, "learning_rate": 2.133985094959532e-05, "loss": 0.0009, "step": 17080 }, { "epoch": 2.46, "learning_rate": 2.1329834121323826e-05, "loss": 0.0007, "step": 17085 }, { "epoch": 2.47, "learning_rate": 2.131981729305233e-05, "loss": 0.0008, "step": 17090 }, { "epoch": 2.47, "learning_rate": 2.1309800464780833e-05, "loss": 0.0007, "step": 17095 }, { "epoch": 2.47, "learning_rate": 2.1299783636509338e-05, "loss": 0.0009, "step": 17100 }, { "epoch": 2.47, "learning_rate": 2.128976680823784e-05, "loss": 0.0006, "step": 17105 }, { "epoch": 2.47, "learning_rate": 2.1279749979966345e-05, "loss": 0.0009, "step": 17110 }, { "epoch": 2.47, "learning_rate": 2.1269733151694847e-05, "loss": 0.0012, "step": 17115 }, { "epoch": 2.47, "learning_rate": 2.1259716323423352e-05, "loss": 0.0008, "step": 17120 }, { "epoch": 2.47, "learning_rate": 2.1249699495151857e-05, "loss": 0.0007, "step": 17125 }, { "epoch": 2.47, "learning_rate": 2.123968266688036e-05, "loss": 0.0008, "step": 17130 }, { "epoch": 2.47, "learning_rate": 2.1229665838608864e-05, "loss": 0.001, "step": 17135 }, { "epoch": 2.47, "learning_rate": 2.1219649010337366e-05, "loss": 0.0007, "step": 17140 }, { "epoch": 2.47, "learning_rate": 2.120963218206587e-05, "loss": 0.0008, "step": 17145 }, { "epoch": 2.47, "learning_rate": 2.1199615353794373e-05, "loss": 0.0006, "step": 17150 }, { "epoch": 2.47, "learning_rate": 2.118959852552288e-05, "loss": 0.0007, "step": 17155 }, { "epoch": 2.48, "learning_rate": 2.1179581697251384e-05, "loss": 0.0008, "step": 17160 }, { "epoch": 2.48, "learning_rate": 2.116956486897989e-05, "loss": 0.0008, "step": 17165 }, { "epoch": 2.48, "learning_rate": 2.115954804070839e-05, "loss": 0.0006, "step": 17170 }, { "epoch": 2.48, "learning_rate": 2.1149531212436896e-05, "loss": 0.0007, "step": 17175 }, { "epoch": 2.48, "learning_rate": 2.1139514384165398e-05, "loss": 0.0006, "step": 17180 }, { "epoch": 2.48, "learning_rate": 2.1129497555893903e-05, "loss": 0.0008, "step": 17185 }, { "epoch": 2.48, "learning_rate": 2.1119480727622405e-05, "loss": 0.0006, "step": 17190 }, { "epoch": 2.48, "learning_rate": 2.110946389935091e-05, "loss": 0.0004, "step": 17195 }, { "epoch": 2.48, "learning_rate": 2.1099447071079415e-05, "loss": 0.0006, "step": 17200 }, { "epoch": 2.48, "learning_rate": 2.1089430242807917e-05, "loss": 0.001, "step": 17205 }, { "epoch": 2.48, "learning_rate": 2.1079413414536422e-05, "loss": 0.0009, "step": 17210 }, { "epoch": 2.48, "learning_rate": 2.1069396586264924e-05, "loss": 0.0007, "step": 17215 }, { "epoch": 2.48, "learning_rate": 2.105937975799343e-05, "loss": 0.0006, "step": 17220 }, { "epoch": 2.48, "learning_rate": 2.104936292972193e-05, "loss": 0.0005, "step": 17225 }, { "epoch": 2.49, "learning_rate": 2.103934610145044e-05, "loss": 0.0012, "step": 17230 }, { "epoch": 2.49, "learning_rate": 2.1029329273178942e-05, "loss": 0.0011, "step": 17235 }, { "epoch": 2.49, "learning_rate": 2.1019312444907447e-05, "loss": 0.0009, "step": 17240 }, { "epoch": 2.49, "learning_rate": 2.100929561663595e-05, "loss": 0.001, "step": 17245 }, { "epoch": 2.49, "learning_rate": 2.0999278788364454e-05, "loss": 0.0005, "step": 17250 }, { "epoch": 2.49, "learning_rate": 2.0989261960092956e-05, "loss": 0.0011, "step": 17255 }, { "epoch": 2.49, "learning_rate": 2.097924513182146e-05, "loss": 0.0009, "step": 17260 }, { "epoch": 2.49, "learning_rate": 2.0969228303549966e-05, "loss": 0.0008, "step": 17265 }, { "epoch": 2.49, "learning_rate": 2.095921147527847e-05, "loss": 0.0007, "step": 17270 }, { "epoch": 2.49, "learning_rate": 2.0949194647006973e-05, "loss": 0.0009, "step": 17275 }, { "epoch": 2.49, "learning_rate": 2.0939177818735475e-05, "loss": 0.0006, "step": 17280 }, { "epoch": 2.49, "learning_rate": 2.092916099046398e-05, "loss": 0.0006, "step": 17285 }, { "epoch": 2.49, "learning_rate": 2.0919144162192482e-05, "loss": 0.0008, "step": 17290 }, { "epoch": 2.49, "learning_rate": 2.0909127333920988e-05, "loss": 0.0004, "step": 17295 }, { "epoch": 2.5, "learning_rate": 2.0899110505649493e-05, "loss": 0.0006, "step": 17300 }, { "epoch": 2.5, "learning_rate": 2.0889093677377998e-05, "loss": 0.001, "step": 17305 }, { "epoch": 2.5, "learning_rate": 2.08790768491065e-05, "loss": 0.0009, "step": 17310 }, { "epoch": 2.5, "learning_rate": 2.0869060020835005e-05, "loss": 0.0009, "step": 17315 }, { "epoch": 2.5, "learning_rate": 2.0859043192563507e-05, "loss": 0.0009, "step": 17320 }, { "epoch": 2.5, "learning_rate": 2.0849026364292012e-05, "loss": 0.0009, "step": 17325 }, { "epoch": 2.5, "learning_rate": 2.0839009536020514e-05, "loss": 0.001, "step": 17330 }, { "epoch": 2.5, "learning_rate": 2.082899270774902e-05, "loss": 0.0009, "step": 17335 }, { "epoch": 2.5, "learning_rate": 2.0818975879477524e-05, "loss": 0.0008, "step": 17340 }, { "epoch": 2.5, "learning_rate": 2.080895905120603e-05, "loss": 0.0009, "step": 17345 }, { "epoch": 2.5, "learning_rate": 2.079894222293453e-05, "loss": 0.0006, "step": 17350 }, { "epoch": 2.5, "learning_rate": 2.0788925394663033e-05, "loss": 0.0008, "step": 17355 }, { "epoch": 2.5, "learning_rate": 2.077890856639154e-05, "loss": 0.0006, "step": 17360 }, { "epoch": 2.5, "learning_rate": 2.076889173812004e-05, "loss": 0.0006, "step": 17365 }, { "epoch": 2.51, "learning_rate": 2.0758874909848546e-05, "loss": 0.0007, "step": 17370 }, { "epoch": 2.51, "learning_rate": 2.074885808157705e-05, "loss": 0.0008, "step": 17375 }, { "epoch": 2.51, "learning_rate": 2.0738841253305556e-05, "loss": 0.0009, "step": 17380 }, { "epoch": 2.51, "learning_rate": 2.0728824425034058e-05, "loss": 0.0012, "step": 17385 }, { "epoch": 2.51, "learning_rate": 2.0718807596762563e-05, "loss": 0.0008, "step": 17390 }, { "epoch": 2.51, "learning_rate": 2.0708790768491065e-05, "loss": 0.0007, "step": 17395 }, { "epoch": 2.51, "learning_rate": 2.069877394021957e-05, "loss": 0.0008, "step": 17400 }, { "epoch": 2.51, "learning_rate": 2.0688757111948072e-05, "loss": 0.0008, "step": 17405 }, { "epoch": 2.51, "learning_rate": 2.0678740283676577e-05, "loss": 0.001, "step": 17410 }, { "epoch": 2.51, "learning_rate": 2.0668723455405083e-05, "loss": 0.0008, "step": 17415 }, { "epoch": 2.51, "learning_rate": 2.0658706627133588e-05, "loss": 0.0012, "step": 17420 }, { "epoch": 2.51, "learning_rate": 2.064868979886209e-05, "loss": 0.0008, "step": 17425 }, { "epoch": 2.51, "learning_rate": 2.0638672970590595e-05, "loss": 0.0009, "step": 17430 }, { "epoch": 2.51, "learning_rate": 2.0628656142319097e-05, "loss": 0.001, "step": 17435 }, { "epoch": 2.52, "learning_rate": 2.06186393140476e-05, "loss": 0.0011, "step": 17440 }, { "epoch": 2.52, "learning_rate": 2.0608622485776104e-05, "loss": 0.0009, "step": 17445 }, { "epoch": 2.52, "learning_rate": 2.059860565750461e-05, "loss": 0.0006, "step": 17450 }, { "epoch": 2.52, "learning_rate": 2.0588588829233114e-05, "loss": 0.001, "step": 17455 }, { "epoch": 2.52, "learning_rate": 2.0578572000961616e-05, "loss": 0.0006, "step": 17460 }, { "epoch": 2.52, "learning_rate": 2.056855517269012e-05, "loss": 0.0009, "step": 17465 }, { "epoch": 2.52, "learning_rate": 2.0558538344418623e-05, "loss": 0.0008, "step": 17470 }, { "epoch": 2.52, "learning_rate": 2.054852151614713e-05, "loss": 0.0008, "step": 17475 }, { "epoch": 2.52, "learning_rate": 2.053850468787563e-05, "loss": 0.0007, "step": 17480 }, { "epoch": 2.52, "learning_rate": 2.0528487859604135e-05, "loss": 0.0006, "step": 17485 }, { "epoch": 2.52, "learning_rate": 2.051847103133264e-05, "loss": 0.0011, "step": 17490 }, { "epoch": 2.52, "learning_rate": 2.0508454203061146e-05, "loss": 0.0009, "step": 17495 }, { "epoch": 2.52, "learning_rate": 2.0498437374789648e-05, "loss": 0.0008, "step": 17500 }, { "epoch": 2.52, "learning_rate": 2.0488420546518153e-05, "loss": 0.0005, "step": 17505 }, { "epoch": 2.53, "learning_rate": 2.0478403718246655e-05, "loss": 0.0006, "step": 17510 }, { "epoch": 2.53, "learning_rate": 2.0468386889975157e-05, "loss": 0.001, "step": 17515 }, { "epoch": 2.53, "learning_rate": 2.0458370061703662e-05, "loss": 0.0009, "step": 17520 }, { "epoch": 2.53, "learning_rate": 2.0448353233432167e-05, "loss": 0.0011, "step": 17525 }, { "epoch": 2.53, "learning_rate": 2.0438336405160672e-05, "loss": 0.0009, "step": 17530 }, { "epoch": 2.53, "learning_rate": 2.0428319576889174e-05, "loss": 0.0006, "step": 17535 }, { "epoch": 2.53, "learning_rate": 2.041830274861768e-05, "loss": 0.0009, "step": 17540 }, { "epoch": 2.53, "learning_rate": 2.040828592034618e-05, "loss": 0.0007, "step": 17545 }, { "epoch": 2.53, "learning_rate": 2.0398269092074686e-05, "loss": 0.0006, "step": 17550 }, { "epoch": 2.53, "learning_rate": 2.0388252263803188e-05, "loss": 0.001, "step": 17555 }, { "epoch": 2.53, "learning_rate": 2.0378235435531693e-05, "loss": 0.0009, "step": 17560 }, { "epoch": 2.53, "learning_rate": 2.03682186072602e-05, "loss": 0.0007, "step": 17565 }, { "epoch": 2.53, "learning_rate": 2.0358201778988704e-05, "loss": 0.0008, "step": 17570 }, { "epoch": 2.53, "learning_rate": 2.0348184950717206e-05, "loss": 0.0009, "step": 17575 }, { "epoch": 2.54, "learning_rate": 2.033816812244571e-05, "loss": 0.001, "step": 17580 }, { "epoch": 2.54, "learning_rate": 2.0328151294174213e-05, "loss": 0.0008, "step": 17585 }, { "epoch": 2.54, "learning_rate": 2.0318134465902715e-05, "loss": 0.0012, "step": 17590 }, { "epoch": 2.54, "learning_rate": 2.030811763763122e-05, "loss": 0.0007, "step": 17595 }, { "epoch": 2.54, "learning_rate": 2.0298100809359725e-05, "loss": 0.0007, "step": 17600 }, { "epoch": 2.54, "learning_rate": 2.028808398108823e-05, "loss": 0.0007, "step": 17605 }, { "epoch": 2.54, "learning_rate": 2.0278067152816732e-05, "loss": 0.0009, "step": 17610 }, { "epoch": 2.54, "learning_rate": 2.0268050324545237e-05, "loss": 0.0005, "step": 17615 }, { "epoch": 2.54, "learning_rate": 2.025803349627374e-05, "loss": 0.0007, "step": 17620 }, { "epoch": 2.54, "learning_rate": 2.0248016668002244e-05, "loss": 0.0011, "step": 17625 }, { "epoch": 2.54, "learning_rate": 2.0237999839730746e-05, "loss": 0.0008, "step": 17630 }, { "epoch": 2.54, "learning_rate": 2.0227983011459255e-05, "loss": 0.0009, "step": 17635 }, { "epoch": 2.54, "learning_rate": 2.0217966183187757e-05, "loss": 0.0008, "step": 17640 }, { "epoch": 2.55, "learning_rate": 2.0207949354916262e-05, "loss": 0.0016, "step": 17645 }, { "epoch": 2.55, "learning_rate": 2.0197932526644764e-05, "loss": 0.0009, "step": 17650 }, { "epoch": 2.55, "learning_rate": 2.018791569837327e-05, "loss": 0.0007, "step": 17655 }, { "epoch": 2.55, "learning_rate": 2.017789887010177e-05, "loss": 0.0009, "step": 17660 }, { "epoch": 2.55, "learning_rate": 2.0167882041830276e-05, "loss": 0.0011, "step": 17665 }, { "epoch": 2.55, "learning_rate": 2.015786521355878e-05, "loss": 0.0007, "step": 17670 }, { "epoch": 2.55, "learning_rate": 2.0147848385287283e-05, "loss": 0.0009, "step": 17675 }, { "epoch": 2.55, "learning_rate": 2.013783155701579e-05, "loss": 0.0011, "step": 17680 }, { "epoch": 2.55, "learning_rate": 2.012781472874429e-05, "loss": 0.0005, "step": 17685 }, { "epoch": 2.55, "learning_rate": 2.0117797900472795e-05, "loss": 0.0009, "step": 17690 }, { "epoch": 2.55, "learning_rate": 2.0107781072201297e-05, "loss": 0.0007, "step": 17695 }, { "epoch": 2.55, "learning_rate": 2.0097764243929803e-05, "loss": 0.0009, "step": 17700 }, { "epoch": 2.55, "learning_rate": 2.0087747415658308e-05, "loss": 0.0008, "step": 17705 }, { "epoch": 2.55, "learning_rate": 2.0077730587386813e-05, "loss": 0.0007, "step": 17710 }, { "epoch": 2.56, "learning_rate": 2.0067713759115315e-05, "loss": 0.001, "step": 17715 }, { "epoch": 2.56, "learning_rate": 2.005769693084382e-05, "loss": 0.001, "step": 17720 }, { "epoch": 2.56, "learning_rate": 2.0047680102572322e-05, "loss": 0.0014, "step": 17725 }, { "epoch": 2.56, "learning_rate": 2.0037663274300827e-05, "loss": 0.0008, "step": 17730 }, { "epoch": 2.56, "learning_rate": 2.002764644602933e-05, "loss": 0.001, "step": 17735 }, { "epoch": 2.56, "learning_rate": 2.0017629617757834e-05, "loss": 0.0008, "step": 17740 }, { "epoch": 2.56, "learning_rate": 2.000761278948634e-05, "loss": 0.0007, "step": 17745 }, { "epoch": 2.56, "learning_rate": 1.999759596121484e-05, "loss": 0.0008, "step": 17750 }, { "epoch": 2.56, "learning_rate": 1.9987579132943346e-05, "loss": 0.0012, "step": 17755 }, { "epoch": 2.56, "learning_rate": 1.997756230467185e-05, "loss": 0.0009, "step": 17760 }, { "epoch": 2.56, "learning_rate": 1.9967545476400354e-05, "loss": 0.0008, "step": 17765 }, { "epoch": 2.56, "learning_rate": 1.9957528648128855e-05, "loss": 0.0005, "step": 17770 }, { "epoch": 2.56, "learning_rate": 1.994751181985736e-05, "loss": 0.0009, "step": 17775 }, { "epoch": 2.56, "learning_rate": 1.9937494991585866e-05, "loss": 0.0012, "step": 17780 }, { "epoch": 2.57, "learning_rate": 1.992747816331437e-05, "loss": 0.0007, "step": 17785 }, { "epoch": 2.57, "learning_rate": 1.9917461335042873e-05, "loss": 0.0006, "step": 17790 }, { "epoch": 2.57, "learning_rate": 1.9907444506771378e-05, "loss": 0.0009, "step": 17795 }, { "epoch": 2.57, "learning_rate": 1.989742767849988e-05, "loss": 0.0008, "step": 17800 }, { "epoch": 2.57, "learning_rate": 1.9887410850228385e-05, "loss": 0.0007, "step": 17805 }, { "epoch": 2.57, "learning_rate": 1.9877394021956887e-05, "loss": 0.001, "step": 17810 }, { "epoch": 2.57, "learning_rate": 1.9867377193685392e-05, "loss": 0.0005, "step": 17815 }, { "epoch": 2.57, "learning_rate": 1.9857360365413897e-05, "loss": 0.0011, "step": 17820 }, { "epoch": 2.57, "learning_rate": 1.98473435371424e-05, "loss": 0.0007, "step": 17825 }, { "epoch": 2.57, "learning_rate": 1.9837326708870905e-05, "loss": 0.0009, "step": 17830 }, { "epoch": 2.57, "learning_rate": 1.9827309880599406e-05, "loss": 0.0007, "step": 17835 }, { "epoch": 2.57, "learning_rate": 1.981729305232791e-05, "loss": 0.0005, "step": 17840 }, { "epoch": 2.57, "learning_rate": 1.9807276224056413e-05, "loss": 0.0009, "step": 17845 }, { "epoch": 2.57, "learning_rate": 1.979725939578492e-05, "loss": 0.0005, "step": 17850 }, { "epoch": 2.58, "learning_rate": 1.9787242567513424e-05, "loss": 0.0011, "step": 17855 }, { "epoch": 2.58, "learning_rate": 1.977722573924193e-05, "loss": 0.0007, "step": 17860 }, { "epoch": 2.58, "learning_rate": 1.976720891097043e-05, "loss": 0.0008, "step": 17865 }, { "epoch": 2.58, "learning_rate": 1.9757192082698936e-05, "loss": 0.0009, "step": 17870 }, { "epoch": 2.58, "learning_rate": 1.9747175254427438e-05, "loss": 0.0006, "step": 17875 }, { "epoch": 2.58, "learning_rate": 1.9737158426155943e-05, "loss": 0.001, "step": 17880 }, { "epoch": 2.58, "learning_rate": 1.9727141597884445e-05, "loss": 0.0009, "step": 17885 }, { "epoch": 2.58, "learning_rate": 1.971712476961295e-05, "loss": 0.0007, "step": 17890 }, { "epoch": 2.58, "learning_rate": 1.9707107941341456e-05, "loss": 0.0008, "step": 17895 }, { "epoch": 2.58, "learning_rate": 1.969709111306996e-05, "loss": 0.0011, "step": 17900 }, { "epoch": 2.58, "learning_rate": 1.9687074284798463e-05, "loss": 0.001, "step": 17905 }, { "epoch": 2.58, "learning_rate": 1.9677057456526964e-05, "loss": 0.0012, "step": 17910 }, { "epoch": 2.58, "learning_rate": 1.966704062825547e-05, "loss": 0.001, "step": 17915 }, { "epoch": 2.58, "learning_rate": 1.965702379998397e-05, "loss": 0.0009, "step": 17920 }, { "epoch": 2.59, "learning_rate": 1.9647006971712477e-05, "loss": 0.0011, "step": 17925 }, { "epoch": 2.59, "learning_rate": 1.9636990143440982e-05, "loss": 0.0004, "step": 17930 }, { "epoch": 2.59, "learning_rate": 1.9626973315169487e-05, "loss": 0.0005, "step": 17935 }, { "epoch": 2.59, "learning_rate": 1.961695648689799e-05, "loss": 0.0009, "step": 17940 }, { "epoch": 2.59, "learning_rate": 1.9606939658626494e-05, "loss": 0.0006, "step": 17945 }, { "epoch": 2.59, "learning_rate": 1.9596922830354996e-05, "loss": 0.0009, "step": 17950 }, { "epoch": 2.59, "learning_rate": 1.95869060020835e-05, "loss": 0.0009, "step": 17955 }, { "epoch": 2.59, "learning_rate": 1.9576889173812003e-05, "loss": 0.0004, "step": 17960 }, { "epoch": 2.59, "learning_rate": 1.956687234554051e-05, "loss": 0.0005, "step": 17965 }, { "epoch": 2.59, "learning_rate": 1.9556855517269014e-05, "loss": 0.0008, "step": 17970 }, { "epoch": 2.59, "learning_rate": 1.954683868899752e-05, "loss": 0.0008, "step": 17975 }, { "epoch": 2.59, "learning_rate": 1.953682186072602e-05, "loss": 0.0011, "step": 17980 }, { "epoch": 2.59, "learning_rate": 1.9526805032454523e-05, "loss": 0.0009, "step": 17985 }, { "epoch": 2.59, "learning_rate": 1.9516788204183028e-05, "loss": 0.0007, "step": 17990 }, { "epoch": 2.6, "learning_rate": 1.950677137591153e-05, "loss": 0.001, "step": 17995 }, { "epoch": 2.6, "learning_rate": 1.9496754547640035e-05, "loss": 0.0007, "step": 18000 }, { "epoch": 2.6, "learning_rate": 1.948673771936854e-05, "loss": 0.0008, "step": 18005 }, { "epoch": 2.6, "learning_rate": 1.9476720891097045e-05, "loss": 0.0011, "step": 18010 }, { "epoch": 2.6, "learning_rate": 1.9466704062825547e-05, "loss": 0.0005, "step": 18015 }, { "epoch": 2.6, "learning_rate": 1.9456687234554052e-05, "loss": 0.0008, "step": 18020 }, { "epoch": 2.6, "learning_rate": 1.9446670406282554e-05, "loss": 0.0009, "step": 18025 }, { "epoch": 2.6, "learning_rate": 1.943665357801106e-05, "loss": 0.0008, "step": 18030 }, { "epoch": 2.6, "learning_rate": 1.942663674973956e-05, "loss": 0.0009, "step": 18035 }, { "epoch": 2.6, "learning_rate": 1.941661992146807e-05, "loss": 0.001, "step": 18040 }, { "epoch": 2.6, "learning_rate": 1.9406603093196572e-05, "loss": 0.0009, "step": 18045 }, { "epoch": 2.6, "learning_rate": 1.9396586264925077e-05, "loss": 0.001, "step": 18050 }, { "epoch": 2.6, "learning_rate": 1.938656943665358e-05, "loss": 0.0007, "step": 18055 }, { "epoch": 2.6, "learning_rate": 1.9376552608382084e-05, "loss": 0.0011, "step": 18060 }, { "epoch": 2.61, "learning_rate": 1.9366535780110586e-05, "loss": 0.0008, "step": 18065 }, { "epoch": 2.61, "learning_rate": 1.9356518951839088e-05, "loss": 0.0009, "step": 18070 }, { "epoch": 2.61, "learning_rate": 1.9346502123567596e-05, "loss": 0.0006, "step": 18075 }, { "epoch": 2.61, "learning_rate": 1.9336485295296098e-05, "loss": 0.001, "step": 18080 }, { "epoch": 2.61, "learning_rate": 1.9326468467024603e-05, "loss": 0.0006, "step": 18085 }, { "epoch": 2.61, "learning_rate": 1.9316451638753105e-05, "loss": 0.0009, "step": 18090 }, { "epoch": 2.61, "learning_rate": 1.930643481048161e-05, "loss": 0.0009, "step": 18095 }, { "epoch": 2.61, "learning_rate": 1.9296417982210112e-05, "loss": 0.0009, "step": 18100 }, { "epoch": 2.61, "learning_rate": 1.9286401153938617e-05, "loss": 0.0008, "step": 18105 }, { "epoch": 2.61, "learning_rate": 1.9276384325667123e-05, "loss": 0.001, "step": 18110 }, { "epoch": 2.61, "learning_rate": 1.9266367497395628e-05, "loss": 0.0007, "step": 18115 }, { "epoch": 2.61, "learning_rate": 1.925635066912413e-05, "loss": 0.0007, "step": 18120 }, { "epoch": 2.61, "learning_rate": 1.9246333840852635e-05, "loss": 0.0009, "step": 18125 }, { "epoch": 2.62, "learning_rate": 1.9236317012581137e-05, "loss": 0.0008, "step": 18130 }, { "epoch": 2.62, "learning_rate": 1.9226300184309642e-05, "loss": 0.0008, "step": 18135 }, { "epoch": 2.62, "learning_rate": 1.9216283356038144e-05, "loss": 0.0009, "step": 18140 }, { "epoch": 2.62, "learning_rate": 1.920626652776665e-05, "loss": 0.0009, "step": 18145 }, { "epoch": 2.62, "learning_rate": 1.9196249699495154e-05, "loss": 0.0007, "step": 18150 }, { "epoch": 2.62, "learning_rate": 1.9186232871223656e-05, "loss": 0.0008, "step": 18155 }, { "epoch": 2.62, "learning_rate": 1.917621604295216e-05, "loss": 0.0012, "step": 18160 }, { "epoch": 2.62, "learning_rate": 1.9166199214680663e-05, "loss": 0.0009, "step": 18165 }, { "epoch": 2.62, "learning_rate": 1.915618238640917e-05, "loss": 0.0006, "step": 18170 }, { "epoch": 2.62, "learning_rate": 1.914616555813767e-05, "loss": 0.0012, "step": 18175 }, { "epoch": 2.62, "learning_rate": 1.9136148729866176e-05, "loss": 0.0008, "step": 18180 }, { "epoch": 2.62, "learning_rate": 1.912613190159468e-05, "loss": 0.0005, "step": 18185 }, { "epoch": 2.62, "learning_rate": 1.9116115073323186e-05, "loss": 0.0007, "step": 18190 }, { "epoch": 2.62, "learning_rate": 1.9106098245051688e-05, "loss": 0.0009, "step": 18195 }, { "epoch": 2.63, "learning_rate": 1.9096081416780193e-05, "loss": 0.001, "step": 18200 }, { "epoch": 2.63, "learning_rate": 1.9086064588508695e-05, "loss": 0.0008, "step": 18205 }, { "epoch": 2.63, "learning_rate": 1.90760477602372e-05, "loss": 0.0007, "step": 18210 }, { "epoch": 2.63, "learning_rate": 1.9066030931965702e-05, "loss": 0.0008, "step": 18215 }, { "epoch": 2.63, "learning_rate": 1.9056014103694207e-05, "loss": 0.0008, "step": 18220 }, { "epoch": 2.63, "learning_rate": 1.9045997275422712e-05, "loss": 0.0008, "step": 18225 }, { "epoch": 2.63, "learning_rate": 1.9035980447151214e-05, "loss": 0.0006, "step": 18230 }, { "epoch": 2.63, "learning_rate": 1.902596361887972e-05, "loss": 0.0008, "step": 18235 }, { "epoch": 2.63, "learning_rate": 1.901594679060822e-05, "loss": 0.0009, "step": 18240 }, { "epoch": 2.63, "learning_rate": 1.9005929962336727e-05, "loss": 0.001, "step": 18245 }, { "epoch": 2.63, "learning_rate": 1.899591313406523e-05, "loss": 0.0007, "step": 18250 }, { "epoch": 2.63, "learning_rate": 1.8985896305793734e-05, "loss": 0.0006, "step": 18255 }, { "epoch": 2.63, "learning_rate": 1.897587947752224e-05, "loss": 0.001, "step": 18260 }, { "epoch": 2.63, "learning_rate": 1.8965862649250744e-05, "loss": 0.0006, "step": 18265 }, { "epoch": 2.64, "learning_rate": 1.8955845820979246e-05, "loss": 0.0004, "step": 18270 }, { "epoch": 2.64, "learning_rate": 1.894582899270775e-05, "loss": 0.0005, "step": 18275 }, { "epoch": 2.64, "learning_rate": 1.8935812164436253e-05, "loss": 0.0007, "step": 18280 }, { "epoch": 2.64, "learning_rate": 1.8925795336164758e-05, "loss": 0.001, "step": 18285 }, { "epoch": 2.64, "learning_rate": 1.891577850789326e-05, "loss": 0.0009, "step": 18290 }, { "epoch": 2.64, "learning_rate": 1.8905761679621765e-05, "loss": 0.0008, "step": 18295 }, { "epoch": 2.64, "learning_rate": 1.889574485135027e-05, "loss": 0.0012, "step": 18300 }, { "epoch": 2.64, "learning_rate": 1.8885728023078772e-05, "loss": 0.0006, "step": 18305 }, { "epoch": 2.64, "learning_rate": 1.8875711194807278e-05, "loss": 0.0012, "step": 18310 }, { "epoch": 2.64, "learning_rate": 1.886569436653578e-05, "loss": 0.0006, "step": 18315 }, { "epoch": 2.64, "learning_rate": 1.8855677538264285e-05, "loss": 0.0009, "step": 18320 }, { "epoch": 2.64, "learning_rate": 1.8845660709992786e-05, "loss": 0.0009, "step": 18325 }, { "epoch": 2.64, "learning_rate": 1.8835643881721292e-05, "loss": 0.0006, "step": 18330 }, { "epoch": 2.64, "learning_rate": 1.8825627053449797e-05, "loss": 0.0006, "step": 18335 }, { "epoch": 2.65, "learning_rate": 1.8815610225178302e-05, "loss": 0.0008, "step": 18340 }, { "epoch": 2.65, "learning_rate": 1.8805593396906804e-05, "loss": 0.0005, "step": 18345 }, { "epoch": 2.65, "learning_rate": 1.879557656863531e-05, "loss": 0.0008, "step": 18350 }, { "epoch": 2.65, "learning_rate": 1.878555974036381e-05, "loss": 0.0009, "step": 18355 }, { "epoch": 2.65, "learning_rate": 1.8775542912092316e-05, "loss": 0.001, "step": 18360 }, { "epoch": 2.65, "learning_rate": 1.8765526083820818e-05, "loss": 0.0011, "step": 18365 }, { "epoch": 2.65, "learning_rate": 1.8755509255549323e-05, "loss": 0.0006, "step": 18370 }, { "epoch": 2.65, "learning_rate": 1.874549242727783e-05, "loss": 0.0008, "step": 18375 }, { "epoch": 2.65, "learning_rate": 1.873547559900633e-05, "loss": 0.0007, "step": 18380 }, { "epoch": 2.65, "learning_rate": 1.8725458770734836e-05, "loss": 0.0006, "step": 18385 }, { "epoch": 2.65, "learning_rate": 1.8715441942463337e-05, "loss": 0.0008, "step": 18390 }, { "epoch": 2.65, "learning_rate": 1.8705425114191843e-05, "loss": 0.0012, "step": 18395 }, { "epoch": 2.65, "learning_rate": 1.8695408285920345e-05, "loss": 0.001, "step": 18400 }, { "epoch": 2.65, "learning_rate": 1.868539145764885e-05, "loss": 0.0006, "step": 18405 }, { "epoch": 2.66, "learning_rate": 1.8675374629377355e-05, "loss": 0.0006, "step": 18410 }, { "epoch": 2.66, "learning_rate": 1.866535780110586e-05, "loss": 0.0006, "step": 18415 }, { "epoch": 2.66, "learning_rate": 1.8655340972834362e-05, "loss": 0.0007, "step": 18420 }, { "epoch": 2.66, "learning_rate": 1.8645324144562867e-05, "loss": 0.0006, "step": 18425 }, { "epoch": 2.66, "learning_rate": 1.863530731629137e-05, "loss": 0.0009, "step": 18430 }, { "epoch": 2.66, "learning_rate": 1.8625290488019874e-05, "loss": 0.0013, "step": 18435 }, { "epoch": 2.66, "learning_rate": 1.8615273659748376e-05, "loss": 0.0009, "step": 18440 }, { "epoch": 2.66, "learning_rate": 1.8605256831476885e-05, "loss": 0.0009, "step": 18445 }, { "epoch": 2.66, "learning_rate": 1.8595240003205387e-05, "loss": 0.0007, "step": 18450 }, { "epoch": 2.66, "learning_rate": 1.8585223174933892e-05, "loss": 0.0008, "step": 18455 }, { "epoch": 2.66, "learning_rate": 1.8575206346662394e-05, "loss": 0.0006, "step": 18460 }, { "epoch": 2.66, "learning_rate": 1.8565189518390896e-05, "loss": 0.0009, "step": 18465 }, { "epoch": 2.66, "learning_rate": 1.85551726901194e-05, "loss": 0.0008, "step": 18470 }, { "epoch": 2.66, "learning_rate": 1.8545155861847903e-05, "loss": 0.0006, "step": 18475 }, { "epoch": 2.67, "learning_rate": 1.853513903357641e-05, "loss": 0.0009, "step": 18480 }, { "epoch": 2.67, "learning_rate": 1.8525122205304913e-05, "loss": 0.0009, "step": 18485 }, { "epoch": 2.67, "learning_rate": 1.851510537703342e-05, "loss": 0.0008, "step": 18490 }, { "epoch": 2.67, "learning_rate": 1.850508854876192e-05, "loss": 0.0008, "step": 18495 }, { "epoch": 2.67, "learning_rate": 1.8495071720490425e-05, "loss": 0.0008, "step": 18500 }, { "epoch": 2.67, "learning_rate": 1.8485054892218927e-05, "loss": 0.0009, "step": 18505 }, { "epoch": 2.67, "learning_rate": 1.8475038063947432e-05, "loss": 0.0006, "step": 18510 }, { "epoch": 2.67, "learning_rate": 1.8465021235675938e-05, "loss": 0.0008, "step": 18515 }, { "epoch": 2.67, "learning_rate": 1.8455004407404443e-05, "loss": 0.0014, "step": 18520 }, { "epoch": 2.67, "learning_rate": 1.8444987579132945e-05, "loss": 0.0005, "step": 18525 }, { "epoch": 2.67, "learning_rate": 1.843497075086145e-05, "loss": 0.0009, "step": 18530 }, { "epoch": 2.67, "learning_rate": 1.8424953922589952e-05, "loss": 0.0005, "step": 18535 }, { "epoch": 2.67, "learning_rate": 1.8414937094318454e-05, "loss": 0.0008, "step": 18540 }, { "epoch": 2.67, "learning_rate": 1.840492026604696e-05, "loss": 0.0004, "step": 18545 }, { "epoch": 2.68, "learning_rate": 1.8394903437775464e-05, "loss": 0.0005, "step": 18550 }, { "epoch": 2.68, "learning_rate": 1.838488660950397e-05, "loss": 0.0006, "step": 18555 }, { "epoch": 2.68, "learning_rate": 1.837486978123247e-05, "loss": 0.0007, "step": 18560 }, { "epoch": 2.68, "learning_rate": 1.8364852952960976e-05, "loss": 0.001, "step": 18565 }, { "epoch": 2.68, "learning_rate": 1.8354836124689478e-05, "loss": 0.0009, "step": 18570 }, { "epoch": 2.68, "learning_rate": 1.8344819296417983e-05, "loss": 0.0009, "step": 18575 }, { "epoch": 2.68, "learning_rate": 1.8334802468146485e-05, "loss": 0.001, "step": 18580 }, { "epoch": 2.68, "learning_rate": 1.832478563987499e-05, "loss": 0.0009, "step": 18585 }, { "epoch": 2.68, "learning_rate": 1.8314768811603496e-05, "loss": 0.001, "step": 18590 }, { "epoch": 2.68, "learning_rate": 1.8304751983332e-05, "loss": 0.001, "step": 18595 }, { "epoch": 2.68, "learning_rate": 1.8294735155060503e-05, "loss": 0.001, "step": 18600 }, { "epoch": 2.68, "learning_rate": 1.8284718326789008e-05, "loss": 0.0006, "step": 18605 }, { "epoch": 2.68, "learning_rate": 1.827470149851751e-05, "loss": 0.0008, "step": 18610 }, { "epoch": 2.68, "learning_rate": 1.8264684670246012e-05, "loss": 0.0012, "step": 18615 }, { "epoch": 2.69, "learning_rate": 1.8254667841974517e-05, "loss": 0.001, "step": 18620 }, { "epoch": 2.69, "learning_rate": 1.8244651013703022e-05, "loss": 0.0009, "step": 18625 }, { "epoch": 2.69, "learning_rate": 1.8234634185431527e-05, "loss": 0.0006, "step": 18630 }, { "epoch": 2.69, "learning_rate": 1.822461735716003e-05, "loss": 0.0009, "step": 18635 }, { "epoch": 2.69, "learning_rate": 1.8214600528888534e-05, "loss": 0.0007, "step": 18640 }, { "epoch": 2.69, "learning_rate": 1.8204583700617036e-05, "loss": 0.001, "step": 18645 }, { "epoch": 2.69, "learning_rate": 1.819456687234554e-05, "loss": 0.0008, "step": 18650 }, { "epoch": 2.69, "learning_rate": 1.8184550044074043e-05, "loss": 0.0006, "step": 18655 }, { "epoch": 2.69, "learning_rate": 1.817453321580255e-05, "loss": 0.0005, "step": 18660 }, { "epoch": 2.69, "learning_rate": 1.8164516387531054e-05, "loss": 0.0011, "step": 18665 }, { "epoch": 2.69, "learning_rate": 1.815449955925956e-05, "loss": 0.0007, "step": 18670 }, { "epoch": 2.69, "learning_rate": 1.814448273098806e-05, "loss": 0.0008, "step": 18675 }, { "epoch": 2.69, "learning_rate": 1.8134465902716566e-05, "loss": 0.0011, "step": 18680 }, { "epoch": 2.7, "learning_rate": 1.8124449074445068e-05, "loss": 0.0009, "step": 18685 }, { "epoch": 2.7, "learning_rate": 1.8114432246173573e-05, "loss": 0.0007, "step": 18690 }, { "epoch": 2.7, "learning_rate": 1.8104415417902075e-05, "loss": 0.001, "step": 18695 }, { "epoch": 2.7, "learning_rate": 1.809439858963058e-05, "loss": 0.0006, "step": 18700 }, { "epoch": 2.7, "learning_rate": 1.8084381761359085e-05, "loss": 0.0008, "step": 18705 }, { "epoch": 2.7, "learning_rate": 1.8074364933087587e-05, "loss": 0.0012, "step": 18710 }, { "epoch": 2.7, "learning_rate": 1.8064348104816093e-05, "loss": 0.0006, "step": 18715 }, { "epoch": 2.7, "learning_rate": 1.8054331276544594e-05, "loss": 0.0007, "step": 18720 }, { "epoch": 2.7, "learning_rate": 1.80443144482731e-05, "loss": 0.0008, "step": 18725 }, { "epoch": 2.7, "learning_rate": 1.80342976200016e-05, "loss": 0.001, "step": 18730 }, { "epoch": 2.7, "learning_rate": 1.8024280791730107e-05, "loss": 0.0008, "step": 18735 }, { "epoch": 2.7, "learning_rate": 1.8014263963458612e-05, "loss": 0.0008, "step": 18740 }, { "epoch": 2.7, "learning_rate": 1.8004247135187117e-05, "loss": 0.0007, "step": 18745 }, { "epoch": 2.7, "learning_rate": 1.799423030691562e-05, "loss": 0.0009, "step": 18750 }, { "epoch": 2.71, "learning_rate": 1.7984213478644124e-05, "loss": 0.0011, "step": 18755 }, { "epoch": 2.71, "learning_rate": 1.7974196650372626e-05, "loss": 0.0008, "step": 18760 }, { "epoch": 2.71, "learning_rate": 1.796417982210113e-05, "loss": 0.0007, "step": 18765 }, { "epoch": 2.71, "learning_rate": 1.7954162993829633e-05, "loss": 0.0011, "step": 18770 }, { "epoch": 2.71, "learning_rate": 1.794414616555814e-05, "loss": 0.0006, "step": 18775 }, { "epoch": 2.71, "learning_rate": 1.7934129337286644e-05, "loss": 0.0009, "step": 18780 }, { "epoch": 2.71, "learning_rate": 1.7924112509015145e-05, "loss": 0.0009, "step": 18785 }, { "epoch": 2.71, "learning_rate": 1.791409568074365e-05, "loss": 0.0009, "step": 18790 }, { "epoch": 2.71, "learning_rate": 1.7904078852472152e-05, "loss": 0.0005, "step": 18795 }, { "epoch": 2.71, "learning_rate": 1.7894062024200658e-05, "loss": 0.0008, "step": 18800 }, { "epoch": 2.71, "learning_rate": 1.788404519592916e-05, "loss": 0.0005, "step": 18805 }, { "epoch": 2.71, "learning_rate": 1.7874028367657665e-05, "loss": 0.0006, "step": 18810 }, { "epoch": 2.71, "learning_rate": 1.786401153938617e-05, "loss": 0.0007, "step": 18815 }, { "epoch": 2.71, "learning_rate": 1.7853994711114675e-05, "loss": 0.0012, "step": 18820 }, { "epoch": 2.72, "learning_rate": 1.7843977882843177e-05, "loss": 0.0008, "step": 18825 }, { "epoch": 2.72, "learning_rate": 1.7833961054571682e-05, "loss": 0.0009, "step": 18830 }, { "epoch": 2.72, "learning_rate": 1.7823944226300184e-05, "loss": 0.0006, "step": 18835 }, { "epoch": 2.72, "learning_rate": 1.781392739802869e-05, "loss": 0.0009, "step": 18840 }, { "epoch": 2.72, "learning_rate": 1.780391056975719e-05, "loss": 0.0005, "step": 18845 }, { "epoch": 2.72, "learning_rate": 1.7793893741485696e-05, "loss": 0.0008, "step": 18850 }, { "epoch": 2.72, "learning_rate": 1.77838769132142e-05, "loss": 0.0007, "step": 18855 }, { "epoch": 2.72, "learning_rate": 1.7773860084942703e-05, "loss": 0.0007, "step": 18860 }, { "epoch": 2.72, "learning_rate": 1.776384325667121e-05, "loss": 0.0007, "step": 18865 }, { "epoch": 2.72, "learning_rate": 1.775382642839971e-05, "loss": 0.0012, "step": 18870 }, { "epoch": 2.72, "learning_rate": 1.7743809600128216e-05, "loss": 0.0007, "step": 18875 }, { "epoch": 2.72, "learning_rate": 1.773379277185672e-05, "loss": 0.0007, "step": 18880 }, { "epoch": 2.72, "learning_rate": 1.7723775943585226e-05, "loss": 0.0011, "step": 18885 }, { "epoch": 2.72, "learning_rate": 1.7713759115313728e-05, "loss": 0.0011, "step": 18890 }, { "epoch": 2.73, "learning_rate": 1.7703742287042233e-05, "loss": 0.0008, "step": 18895 }, { "epoch": 2.73, "learning_rate": 1.7693725458770735e-05, "loss": 0.0008, "step": 18900 }, { "epoch": 2.73, "learning_rate": 1.768370863049924e-05, "loss": 0.0009, "step": 18905 }, { "epoch": 2.73, "learning_rate": 1.7673691802227742e-05, "loss": 0.0006, "step": 18910 }, { "epoch": 2.73, "learning_rate": 1.7663674973956247e-05, "loss": 0.001, "step": 18915 }, { "epoch": 2.73, "learning_rate": 1.7653658145684753e-05, "loss": 0.0008, "step": 18920 }, { "epoch": 2.73, "learning_rate": 1.7643641317413258e-05, "loss": 0.001, "step": 18925 }, { "epoch": 2.73, "learning_rate": 1.763362448914176e-05, "loss": 0.0006, "step": 18930 }, { "epoch": 2.73, "learning_rate": 1.762360766087026e-05, "loss": 0.0009, "step": 18935 }, { "epoch": 2.73, "learning_rate": 1.7613590832598767e-05, "loss": 0.0006, "step": 18940 }, { "epoch": 2.73, "learning_rate": 1.760357400432727e-05, "loss": 0.0008, "step": 18945 }, { "epoch": 2.73, "learning_rate": 1.7593557176055774e-05, "loss": 0.0009, "step": 18950 }, { "epoch": 2.73, "learning_rate": 1.758354034778428e-05, "loss": 0.0008, "step": 18955 }, { "epoch": 2.73, "learning_rate": 1.7573523519512784e-05, "loss": 0.0007, "step": 18960 }, { "epoch": 2.74, "learning_rate": 1.7563506691241286e-05, "loss": 0.0008, "step": 18965 }, { "epoch": 2.74, "learning_rate": 1.755348986296979e-05, "loss": 0.0009, "step": 18970 }, { "epoch": 2.74, "learning_rate": 1.7543473034698293e-05, "loss": 0.0008, "step": 18975 }, { "epoch": 2.74, "learning_rate": 1.75334562064268e-05, "loss": 0.001, "step": 18980 }, { "epoch": 2.74, "learning_rate": 1.75234393781553e-05, "loss": 0.001, "step": 18985 }, { "epoch": 2.74, "learning_rate": 1.7513422549883805e-05, "loss": 0.0008, "step": 18990 }, { "epoch": 2.74, "learning_rate": 1.750340572161231e-05, "loss": 0.0005, "step": 18995 }, { "epoch": 2.74, "learning_rate": 1.7493388893340816e-05, "loss": 0.0008, "step": 19000 }, { "epoch": 2.74, "learning_rate": 1.7483372065069318e-05, "loss": 0.0003, "step": 19005 }, { "epoch": 2.74, "learning_rate": 1.747335523679782e-05, "loss": 0.0009, "step": 19010 }, { "epoch": 2.74, "learning_rate": 1.7463338408526325e-05, "loss": 0.0006, "step": 19015 }, { "epoch": 2.74, "learning_rate": 1.7453321580254827e-05, "loss": 0.0008, "step": 19020 }, { "epoch": 2.74, "learning_rate": 1.7443304751983332e-05, "loss": 0.0007, "step": 19025 }, { "epoch": 2.74, "learning_rate": 1.7433287923711837e-05, "loss": 0.001, "step": 19030 }, { "epoch": 2.75, "learning_rate": 1.7423271095440342e-05, "loss": 0.001, "step": 19035 }, { "epoch": 2.75, "learning_rate": 1.7413254267168844e-05, "loss": 0.001, "step": 19040 }, { "epoch": 2.75, "learning_rate": 1.740323743889735e-05, "loss": 0.0006, "step": 19045 }, { "epoch": 2.75, "learning_rate": 1.739322061062585e-05, "loss": 0.0009, "step": 19050 }, { "epoch": 2.75, "learning_rate": 1.7383203782354356e-05, "loss": 0.0008, "step": 19055 }, { "epoch": 2.75, "learning_rate": 1.737318695408286e-05, "loss": 0.0005, "step": 19060 }, { "epoch": 2.75, "learning_rate": 1.7363170125811364e-05, "loss": 0.0004, "step": 19065 }, { "epoch": 2.75, "learning_rate": 1.735315329753987e-05, "loss": 0.0011, "step": 19070 }, { "epoch": 2.75, "learning_rate": 1.7343136469268374e-05, "loss": 0.0008, "step": 19075 }, { "epoch": 2.75, "learning_rate": 1.7333119640996876e-05, "loss": 0.0008, "step": 19080 }, { "epoch": 2.75, "learning_rate": 1.732310281272538e-05, "loss": 0.001, "step": 19085 }, { "epoch": 2.75, "learning_rate": 1.7313085984453883e-05, "loss": 0.0008, "step": 19090 }, { "epoch": 2.75, "learning_rate": 1.7303069156182385e-05, "loss": 0.0008, "step": 19095 }, { "epoch": 2.75, "learning_rate": 1.729305232791089e-05, "loss": 0.0008, "step": 19100 }, { "epoch": 2.76, "learning_rate": 1.7283035499639395e-05, "loss": 0.0008, "step": 19105 }, { "epoch": 2.76, "learning_rate": 1.72730186713679e-05, "loss": 0.0005, "step": 19110 }, { "epoch": 2.76, "learning_rate": 1.7263001843096402e-05, "loss": 0.0009, "step": 19115 }, { "epoch": 2.76, "learning_rate": 1.7252985014824907e-05, "loss": 0.0008, "step": 19120 }, { "epoch": 2.76, "learning_rate": 1.724296818655341e-05, "loss": 0.0008, "step": 19125 }, { "epoch": 2.76, "learning_rate": 1.7232951358281915e-05, "loss": 0.0005, "step": 19130 }, { "epoch": 2.76, "learning_rate": 1.7222934530010416e-05, "loss": 0.0006, "step": 19135 }, { "epoch": 2.76, "learning_rate": 1.721291770173892e-05, "loss": 0.0008, "step": 19140 }, { "epoch": 2.76, "learning_rate": 1.7202900873467427e-05, "loss": 0.0007, "step": 19145 }, { "epoch": 2.76, "learning_rate": 1.7192884045195932e-05, "loss": 0.0011, "step": 19150 }, { "epoch": 2.76, "learning_rate": 1.7182867216924434e-05, "loss": 0.0005, "step": 19155 }, { "epoch": 2.76, "learning_rate": 1.717285038865294e-05, "loss": 0.0007, "step": 19160 }, { "epoch": 2.76, "learning_rate": 1.716283356038144e-05, "loss": 0.0012, "step": 19165 }, { "epoch": 2.77, "learning_rate": 1.7152816732109943e-05, "loss": 0.0007, "step": 19170 }, { "epoch": 2.77, "learning_rate": 1.7142799903838448e-05, "loss": 0.001, "step": 19175 }, { "epoch": 2.77, "learning_rate": 1.7132783075566953e-05, "loss": 0.0007, "step": 19180 }, { "epoch": 2.77, "learning_rate": 1.712276624729546e-05, "loss": 0.0009, "step": 19185 }, { "epoch": 2.77, "learning_rate": 1.711274941902396e-05, "loss": 0.0008, "step": 19190 }, { "epoch": 2.77, "learning_rate": 1.7102732590752466e-05, "loss": 0.0008, "step": 19195 }, { "epoch": 2.77, "learning_rate": 1.7092715762480967e-05, "loss": 0.0008, "step": 19200 }, { "epoch": 2.77, "learning_rate": 1.7082698934209473e-05, "loss": 0.0008, "step": 19205 }, { "epoch": 2.77, "learning_rate": 1.7072682105937974e-05, "loss": 0.0009, "step": 19210 }, { "epoch": 2.77, "learning_rate": 1.7062665277666483e-05, "loss": 0.0007, "step": 19215 }, { "epoch": 2.77, "learning_rate": 1.7052648449394985e-05, "loss": 0.0007, "step": 19220 }, { "epoch": 2.77, "learning_rate": 1.704263162112349e-05, "loss": 0.001, "step": 19225 }, { "epoch": 2.77, "learning_rate": 1.7032614792851992e-05, "loss": 0.001, "step": 19230 }, { "epoch": 2.77, "learning_rate": 1.7022597964580497e-05, "loss": 0.0006, "step": 19235 }, { "epoch": 2.78, "learning_rate": 1.7012581136309e-05, "loss": 0.0008, "step": 19240 }, { "epoch": 2.78, "learning_rate": 1.70025643080375e-05, "loss": 0.0011, "step": 19245 }, { "epoch": 2.78, "learning_rate": 1.699254747976601e-05, "loss": 0.0009, "step": 19250 }, { "epoch": 2.78, "learning_rate": 1.698253065149451e-05, "loss": 0.0009, "step": 19255 }, { "epoch": 2.78, "learning_rate": 1.6972513823223017e-05, "loss": 0.0009, "step": 19260 }, { "epoch": 2.78, "learning_rate": 1.696249699495152e-05, "loss": 0.0009, "step": 19265 }, { "epoch": 2.78, "learning_rate": 1.6952480166680024e-05, "loss": 0.0005, "step": 19270 }, { "epoch": 2.78, "learning_rate": 1.6942463338408525e-05, "loss": 0.001, "step": 19275 }, { "epoch": 2.78, "learning_rate": 1.693244651013703e-05, "loss": 0.0009, "step": 19280 }, { "epoch": 2.78, "learning_rate": 1.6922429681865536e-05, "loss": 0.0009, "step": 19285 }, { "epoch": 2.78, "learning_rate": 1.691241285359404e-05, "loss": 0.001, "step": 19290 }, { "epoch": 2.78, "learning_rate": 1.6902396025322543e-05, "loss": 0.0005, "step": 19295 }, { "epoch": 2.78, "learning_rate": 1.6892379197051048e-05, "loss": 0.0007, "step": 19300 }, { "epoch": 2.78, "learning_rate": 1.688236236877955e-05, "loss": 0.0006, "step": 19305 }, { "epoch": 2.79, "learning_rate": 1.6872345540508055e-05, "loss": 0.0008, "step": 19310 }, { "epoch": 2.79, "learning_rate": 1.6862328712236557e-05, "loss": 0.0006, "step": 19315 }, { "epoch": 2.79, "learning_rate": 1.6852311883965062e-05, "loss": 0.0008, "step": 19320 }, { "epoch": 2.79, "learning_rate": 1.6842295055693568e-05, "loss": 0.0009, "step": 19325 }, { "epoch": 2.79, "learning_rate": 1.683227822742207e-05, "loss": 0.0008, "step": 19330 }, { "epoch": 2.79, "learning_rate": 1.6822261399150575e-05, "loss": 0.0007, "step": 19335 }, { "epoch": 2.79, "learning_rate": 1.6812244570879076e-05, "loss": 0.001, "step": 19340 }, { "epoch": 2.79, "learning_rate": 1.680222774260758e-05, "loss": 0.0009, "step": 19345 }, { "epoch": 2.79, "learning_rate": 1.6792210914336084e-05, "loss": 0.0007, "step": 19350 }, { "epoch": 2.79, "learning_rate": 1.678219408606459e-05, "loss": 0.0008, "step": 19355 }, { "epoch": 2.79, "learning_rate": 1.6772177257793094e-05, "loss": 0.0007, "step": 19360 }, { "epoch": 2.79, "learning_rate": 1.67621604295216e-05, "loss": 0.0011, "step": 19365 }, { "epoch": 2.79, "learning_rate": 1.67521436012501e-05, "loss": 0.001, "step": 19370 }, { "epoch": 2.79, "learning_rate": 1.6742126772978606e-05, "loss": 0.001, "step": 19375 }, { "epoch": 2.8, "learning_rate": 1.6732109944707108e-05, "loss": 0.0008, "step": 19380 }, { "epoch": 2.8, "learning_rate": 1.6722093116435613e-05, "loss": 0.0008, "step": 19385 }, { "epoch": 2.8, "learning_rate": 1.6712076288164115e-05, "loss": 0.0007, "step": 19390 }, { "epoch": 2.8, "learning_rate": 1.670205945989262e-05, "loss": 0.0009, "step": 19395 }, { "epoch": 2.8, "learning_rate": 1.6692042631621126e-05, "loss": 0.0009, "step": 19400 }, { "epoch": 2.8, "learning_rate": 1.6682025803349627e-05, "loss": 0.0006, "step": 19405 }, { "epoch": 2.8, "learning_rate": 1.6672008975078133e-05, "loss": 0.0007, "step": 19410 }, { "epoch": 2.8, "learning_rate": 1.6661992146806635e-05, "loss": 0.0006, "step": 19415 }, { "epoch": 2.8, "learning_rate": 1.665197531853514e-05, "loss": 0.0009, "step": 19420 }, { "epoch": 2.8, "learning_rate": 1.664195849026364e-05, "loss": 0.0008, "step": 19425 }, { "epoch": 2.8, "learning_rate": 1.6631941661992147e-05, "loss": 0.0006, "step": 19430 }, { "epoch": 2.8, "learning_rate": 1.6621924833720652e-05, "loss": 0.0008, "step": 19435 }, { "epoch": 2.8, "learning_rate": 1.6611908005449157e-05, "loss": 0.0011, "step": 19440 }, { "epoch": 2.8, "learning_rate": 1.660189117717766e-05, "loss": 0.0011, "step": 19445 }, { "epoch": 2.81, "learning_rate": 1.6591874348906164e-05, "loss": 0.0007, "step": 19450 }, { "epoch": 2.81, "learning_rate": 1.6581857520634666e-05, "loss": 0.0006, "step": 19455 }, { "epoch": 2.81, "learning_rate": 1.657184069236317e-05, "loss": 0.0007, "step": 19460 }, { "epoch": 2.81, "learning_rate": 1.6561823864091673e-05, "loss": 0.0008, "step": 19465 }, { "epoch": 2.81, "learning_rate": 1.655180703582018e-05, "loss": 0.001, "step": 19470 }, { "epoch": 2.81, "learning_rate": 1.6541790207548684e-05, "loss": 0.001, "step": 19475 }, { "epoch": 2.81, "learning_rate": 1.653177337927719e-05, "loss": 0.0009, "step": 19480 }, { "epoch": 2.81, "learning_rate": 1.652175655100569e-05, "loss": 0.0008, "step": 19485 }, { "epoch": 2.81, "learning_rate": 1.6511739722734193e-05, "loss": 0.0005, "step": 19490 }, { "epoch": 2.81, "learning_rate": 1.6501722894462698e-05, "loss": 0.0009, "step": 19495 }, { "epoch": 2.81, "learning_rate": 1.64917060661912e-05, "loss": 0.0008, "step": 19500 }, { "epoch": 2.81, "learning_rate": 1.6481689237919705e-05, "loss": 0.0007, "step": 19505 }, { "epoch": 2.81, "learning_rate": 1.647167240964821e-05, "loss": 0.0009, "step": 19510 }, { "epoch": 2.81, "learning_rate": 1.6461655581376715e-05, "loss": 0.0009, "step": 19515 }, { "epoch": 2.82, "learning_rate": 1.6451638753105217e-05, "loss": 0.0006, "step": 19520 }, { "epoch": 2.82, "learning_rate": 1.6441621924833722e-05, "loss": 0.0007, "step": 19525 }, { "epoch": 2.82, "learning_rate": 1.6431605096562224e-05, "loss": 0.0005, "step": 19530 }, { "epoch": 2.82, "learning_rate": 1.642158826829073e-05, "loss": 0.0009, "step": 19535 }, { "epoch": 2.82, "learning_rate": 1.641157144001923e-05, "loss": 0.0009, "step": 19540 }, { "epoch": 2.82, "learning_rate": 1.6401554611747737e-05, "loss": 0.0008, "step": 19545 }, { "epoch": 2.82, "learning_rate": 1.6391537783476242e-05, "loss": 0.0006, "step": 19550 }, { "epoch": 2.82, "learning_rate": 1.6381520955204747e-05, "loss": 0.0008, "step": 19555 }, { "epoch": 2.82, "learning_rate": 1.637150412693325e-05, "loss": 0.0012, "step": 19560 }, { "epoch": 2.82, "learning_rate": 1.636148729866175e-05, "loss": 0.0005, "step": 19565 }, { "epoch": 2.82, "learning_rate": 1.6351470470390256e-05, "loss": 0.0009, "step": 19570 }, { "epoch": 2.82, "learning_rate": 1.6341453642118758e-05, "loss": 0.001, "step": 19575 }, { "epoch": 2.82, "learning_rate": 1.6331436813847263e-05, "loss": 0.0007, "step": 19580 }, { "epoch": 2.82, "learning_rate": 1.6321419985575768e-05, "loss": 0.0006, "step": 19585 }, { "epoch": 2.83, "learning_rate": 1.6311403157304273e-05, "loss": 0.001, "step": 19590 }, { "epoch": 2.83, "learning_rate": 1.6301386329032775e-05, "loss": 0.0005, "step": 19595 }, { "epoch": 2.83, "learning_rate": 1.629136950076128e-05, "loss": 0.0007, "step": 19600 }, { "epoch": 2.83, "learning_rate": 1.6281352672489782e-05, "loss": 0.001, "step": 19605 }, { "epoch": 2.83, "learning_rate": 1.6271335844218288e-05, "loss": 0.0007, "step": 19610 }, { "epoch": 2.83, "learning_rate": 1.626131901594679e-05, "loss": 0.0007, "step": 19615 }, { "epoch": 2.83, "learning_rate": 1.6251302187675298e-05, "loss": 0.0009, "step": 19620 }, { "epoch": 2.83, "learning_rate": 1.62412853594038e-05, "loss": 0.0009, "step": 19625 }, { "epoch": 2.83, "learning_rate": 1.6231268531132305e-05, "loss": 0.0009, "step": 19630 }, { "epoch": 2.83, "learning_rate": 1.6221251702860807e-05, "loss": 0.001, "step": 19635 }, { "epoch": 2.83, "learning_rate": 1.621123487458931e-05, "loss": 0.001, "step": 19640 }, { "epoch": 2.83, "learning_rate": 1.6201218046317814e-05, "loss": 0.0007, "step": 19645 }, { "epoch": 2.83, "learning_rate": 1.6191201218046316e-05, "loss": 0.0009, "step": 19650 }, { "epoch": 2.83, "learning_rate": 1.6181184389774824e-05, "loss": 0.0006, "step": 19655 }, { "epoch": 2.84, "learning_rate": 1.6171167561503326e-05, "loss": 0.0005, "step": 19660 }, { "epoch": 2.84, "learning_rate": 1.616115073323183e-05, "loss": 0.0008, "step": 19665 }, { "epoch": 2.84, "learning_rate": 1.6151133904960333e-05, "loss": 0.0009, "step": 19670 }, { "epoch": 2.84, "learning_rate": 1.614111707668884e-05, "loss": 0.0009, "step": 19675 }, { "epoch": 2.84, "learning_rate": 1.613110024841734e-05, "loss": 0.0006, "step": 19680 }, { "epoch": 2.84, "learning_rate": 1.6121083420145846e-05, "loss": 0.0006, "step": 19685 }, { "epoch": 2.84, "learning_rate": 1.611106659187435e-05, "loss": 0.001, "step": 19690 }, { "epoch": 2.84, "learning_rate": 1.6101049763602856e-05, "loss": 0.0008, "step": 19695 }, { "epoch": 2.84, "learning_rate": 1.6091032935331358e-05, "loss": 0.0006, "step": 19700 }, { "epoch": 2.84, "learning_rate": 1.6081016107059863e-05, "loss": 0.0008, "step": 19705 }, { "epoch": 2.84, "learning_rate": 1.6070999278788365e-05, "loss": 0.0012, "step": 19710 }, { "epoch": 2.84, "learning_rate": 1.606098245051687e-05, "loss": 0.0005, "step": 19715 }, { "epoch": 2.84, "learning_rate": 1.6050965622245372e-05, "loss": 0.0006, "step": 19720 }, { "epoch": 2.85, "learning_rate": 1.6040948793973877e-05, "loss": 0.001, "step": 19725 }, { "epoch": 2.85, "learning_rate": 1.6030931965702383e-05, "loss": 0.0006, "step": 19730 }, { "epoch": 2.85, "learning_rate": 1.6020915137430884e-05, "loss": 0.001, "step": 19735 }, { "epoch": 2.85, "learning_rate": 1.601089830915939e-05, "loss": 0.0014, "step": 19740 }, { "epoch": 2.85, "learning_rate": 1.600088148088789e-05, "loss": 0.0009, "step": 19745 }, { "epoch": 2.85, "learning_rate": 1.5990864652616397e-05, "loss": 0.0007, "step": 19750 }, { "epoch": 2.85, "learning_rate": 1.59808478243449e-05, "loss": 0.0008, "step": 19755 }, { "epoch": 2.85, "learning_rate": 1.5970830996073404e-05, "loss": 0.0008, "step": 19760 }, { "epoch": 2.85, "learning_rate": 1.596081416780191e-05, "loss": 0.0008, "step": 19765 }, { "epoch": 2.85, "learning_rate": 1.5950797339530414e-05, "loss": 0.0009, "step": 19770 }, { "epoch": 2.85, "learning_rate": 1.5940780511258916e-05, "loss": 0.0006, "step": 19775 }, { "epoch": 2.85, "learning_rate": 1.593076368298742e-05, "loss": 0.0007, "step": 19780 }, { "epoch": 2.85, "learning_rate": 1.5920746854715923e-05, "loss": 0.001, "step": 19785 }, { "epoch": 2.85, "learning_rate": 1.5910730026444428e-05, "loss": 0.0007, "step": 19790 }, { "epoch": 2.86, "learning_rate": 1.590071319817293e-05, "loss": 0.0006, "step": 19795 }, { "epoch": 2.86, "learning_rate": 1.5890696369901435e-05, "loss": 0.0009, "step": 19800 }, { "epoch": 2.86, "learning_rate": 1.588067954162994e-05, "loss": 0.0008, "step": 19805 }, { "epoch": 2.86, "learning_rate": 1.5870662713358442e-05, "loss": 0.0011, "step": 19810 }, { "epoch": 2.86, "learning_rate": 1.5860645885086948e-05, "loss": 0.0009, "step": 19815 }, { "epoch": 2.86, "learning_rate": 1.585062905681545e-05, "loss": 0.0009, "step": 19820 }, { "epoch": 2.86, "learning_rate": 1.5840612228543955e-05, "loss": 0.0011, "step": 19825 }, { "epoch": 2.86, "learning_rate": 1.5830595400272457e-05, "loss": 0.0008, "step": 19830 }, { "epoch": 2.86, "learning_rate": 1.5820578572000962e-05, "loss": 0.0007, "step": 19835 }, { "epoch": 2.86, "learning_rate": 1.5810561743729467e-05, "loss": 0.0004, "step": 19840 }, { "epoch": 2.86, "learning_rate": 1.5800544915457972e-05, "loss": 0.0007, "step": 19845 }, { "epoch": 2.86, "learning_rate": 1.5790528087186474e-05, "loss": 0.0007, "step": 19850 }, { "epoch": 2.86, "learning_rate": 1.578051125891498e-05, "loss": 0.0009, "step": 19855 }, { "epoch": 2.86, "learning_rate": 1.577049443064348e-05, "loss": 0.0012, "step": 19860 }, { "epoch": 2.87, "learning_rate": 1.5760477602371986e-05, "loss": 0.0006, "step": 19865 }, { "epoch": 2.87, "learning_rate": 1.5750460774100488e-05, "loss": 0.0007, "step": 19870 }, { "epoch": 2.87, "learning_rate": 1.5740443945828993e-05, "loss": 0.0009, "step": 19875 }, { "epoch": 2.87, "learning_rate": 1.57304271175575e-05, "loss": 0.0009, "step": 19880 }, { "epoch": 2.87, "learning_rate": 1.5720410289286e-05, "loss": 0.0008, "step": 19885 }, { "epoch": 2.87, "learning_rate": 1.5710393461014506e-05, "loss": 0.0007, "step": 19890 }, { "epoch": 2.87, "learning_rate": 1.5700376632743008e-05, "loss": 0.0011, "step": 19895 }, { "epoch": 2.87, "learning_rate": 1.5690359804471513e-05, "loss": 0.0006, "step": 19900 }, { "epoch": 2.87, "learning_rate": 1.5680342976200015e-05, "loss": 0.0006, "step": 19905 }, { "epoch": 2.87, "learning_rate": 1.567032614792852e-05, "loss": 0.0006, "step": 19910 }, { "epoch": 2.87, "learning_rate": 1.5660309319657025e-05, "loss": 0.0007, "step": 19915 }, { "epoch": 2.87, "learning_rate": 1.565029249138553e-05, "loss": 0.0005, "step": 19920 }, { "epoch": 2.87, "learning_rate": 1.5640275663114032e-05, "loss": 0.0008, "step": 19925 }, { "epoch": 2.87, "learning_rate": 1.5630258834842537e-05, "loss": 0.0013, "step": 19930 }, { "epoch": 2.88, "learning_rate": 1.562024200657104e-05, "loss": 0.0009, "step": 19935 }, { "epoch": 2.88, "learning_rate": 1.5610225178299544e-05, "loss": 0.001, "step": 19940 }, { "epoch": 2.88, "learning_rate": 1.5600208350028046e-05, "loss": 0.0009, "step": 19945 }, { "epoch": 2.88, "learning_rate": 1.559019152175655e-05, "loss": 0.0005, "step": 19950 }, { "epoch": 2.88, "learning_rate": 1.5580174693485057e-05, "loss": 0.0009, "step": 19955 }, { "epoch": 2.88, "learning_rate": 1.557015786521356e-05, "loss": 0.0009, "step": 19960 }, { "epoch": 2.88, "learning_rate": 1.5560141036942064e-05, "loss": 0.0015, "step": 19965 }, { "epoch": 2.88, "learning_rate": 1.5550124208670566e-05, "loss": 0.0005, "step": 19970 }, { "epoch": 2.88, "learning_rate": 1.554010738039907e-05, "loss": 0.0007, "step": 19975 }, { "epoch": 2.88, "learning_rate": 1.5530090552127573e-05, "loss": 0.0007, "step": 19980 }, { "epoch": 2.88, "learning_rate": 1.5520073723856078e-05, "loss": 0.0008, "step": 19985 }, { "epoch": 2.88, "learning_rate": 1.5510056895584583e-05, "loss": 0.0009, "step": 19990 }, { "epoch": 2.88, "learning_rate": 1.550004006731309e-05, "loss": 0.0008, "step": 19995 }, { "epoch": 2.88, "learning_rate": 1.549002323904159e-05, "loss": 0.0008, "step": 20000 }, { "epoch": 2.89, "learning_rate": 1.5480006410770095e-05, "loss": 0.0011, "step": 20005 }, { "epoch": 2.89, "learning_rate": 1.5469989582498597e-05, "loss": 0.0007, "step": 20010 }, { "epoch": 2.89, "learning_rate": 1.5459972754227103e-05, "loss": 0.0007, "step": 20015 }, { "epoch": 2.89, "learning_rate": 1.5449955925955604e-05, "loss": 0.0007, "step": 20020 }, { "epoch": 2.89, "learning_rate": 1.5439939097684113e-05, "loss": 0.0009, "step": 20025 }, { "epoch": 2.89, "learning_rate": 1.5429922269412615e-05, "loss": 0.0006, "step": 20030 }, { "epoch": 2.89, "learning_rate": 1.5419905441141117e-05, "loss": 0.001, "step": 20035 }, { "epoch": 2.89, "learning_rate": 1.5409888612869622e-05, "loss": 0.0004, "step": 20040 }, { "epoch": 2.89, "learning_rate": 1.5399871784598124e-05, "loss": 0.0004, "step": 20045 }, { "epoch": 2.89, "learning_rate": 1.538985495632663e-05, "loss": 0.0007, "step": 20050 }, { "epoch": 2.89, "learning_rate": 1.537983812805513e-05, "loss": 0.0009, "step": 20055 }, { "epoch": 2.89, "learning_rate": 1.536982129978364e-05, "loss": 0.0009, "step": 20060 }, { "epoch": 2.89, "learning_rate": 1.535980447151214e-05, "loss": 0.0005, "step": 20065 }, { "epoch": 2.89, "learning_rate": 1.5349787643240646e-05, "loss": 0.0008, "step": 20070 }, { "epoch": 2.9, "learning_rate": 1.5339770814969148e-05, "loss": 0.0007, "step": 20075 }, { "epoch": 2.9, "learning_rate": 1.5329753986697654e-05, "loss": 0.0008, "step": 20080 }, { "epoch": 2.9, "learning_rate": 1.5319737158426155e-05, "loss": 0.0012, "step": 20085 }, { "epoch": 2.9, "learning_rate": 1.530972033015466e-05, "loss": 0.0009, "step": 20090 }, { "epoch": 2.9, "learning_rate": 1.5299703501883166e-05, "loss": 0.0007, "step": 20095 }, { "epoch": 2.9, "learning_rate": 1.528968667361167e-05, "loss": 0.0005, "step": 20100 }, { "epoch": 2.9, "learning_rate": 1.5279669845340173e-05, "loss": 0.0006, "step": 20105 }, { "epoch": 2.9, "learning_rate": 1.5269653017068678e-05, "loss": 0.0007, "step": 20110 }, { "epoch": 2.9, "learning_rate": 1.525963618879718e-05, "loss": 0.0007, "step": 20115 }, { "epoch": 2.9, "learning_rate": 1.5249619360525683e-05, "loss": 0.0006, "step": 20120 }, { "epoch": 2.9, "learning_rate": 1.5239602532254189e-05, "loss": 0.0006, "step": 20125 }, { "epoch": 2.9, "learning_rate": 1.522958570398269e-05, "loss": 0.0008, "step": 20130 }, { "epoch": 2.9, "learning_rate": 1.5219568875711196e-05, "loss": 0.0006, "step": 20135 }, { "epoch": 2.9, "learning_rate": 1.52095520474397e-05, "loss": 0.0008, "step": 20140 }, { "epoch": 2.91, "learning_rate": 1.5199535219168205e-05, "loss": 0.0005, "step": 20145 }, { "epoch": 2.91, "learning_rate": 1.5189518390896706e-05, "loss": 0.0012, "step": 20150 }, { "epoch": 2.91, "learning_rate": 1.5179501562625212e-05, "loss": 0.0006, "step": 20155 }, { "epoch": 2.91, "learning_rate": 1.5169484734353715e-05, "loss": 0.0009, "step": 20160 }, { "epoch": 2.91, "learning_rate": 1.515946790608222e-05, "loss": 0.0007, "step": 20165 }, { "epoch": 2.91, "learning_rate": 1.5149451077810722e-05, "loss": 0.0007, "step": 20170 }, { "epoch": 2.91, "learning_rate": 1.5139434249539227e-05, "loss": 0.0006, "step": 20175 }, { "epoch": 2.91, "learning_rate": 1.5129417421267731e-05, "loss": 0.0004, "step": 20180 }, { "epoch": 2.91, "learning_rate": 1.5119400592996236e-05, "loss": 0.0008, "step": 20185 }, { "epoch": 2.91, "learning_rate": 1.5109383764724738e-05, "loss": 0.001, "step": 20190 }, { "epoch": 2.91, "learning_rate": 1.5099366936453242e-05, "loss": 0.0008, "step": 20195 }, { "epoch": 2.91, "learning_rate": 1.5089350108181747e-05, "loss": 0.0012, "step": 20200 }, { "epoch": 2.91, "learning_rate": 1.5079333279910249e-05, "loss": 0.0008, "step": 20205 }, { "epoch": 2.92, "learning_rate": 1.5069316451638754e-05, "loss": 0.0008, "step": 20210 }, { "epoch": 2.92, "learning_rate": 1.5059299623367257e-05, "loss": 0.0009, "step": 20215 }, { "epoch": 2.92, "learning_rate": 1.5049282795095763e-05, "loss": 0.0008, "step": 20220 }, { "epoch": 2.92, "learning_rate": 1.5039265966824264e-05, "loss": 0.0008, "step": 20225 }, { "epoch": 2.92, "learning_rate": 1.502924913855277e-05, "loss": 0.0006, "step": 20230 }, { "epoch": 2.92, "learning_rate": 1.5019232310281273e-05, "loss": 0.001, "step": 20235 }, { "epoch": 2.92, "learning_rate": 1.5009215482009778e-05, "loss": 0.0006, "step": 20240 }, { "epoch": 2.92, "learning_rate": 1.499919865373828e-05, "loss": 0.0008, "step": 20245 }, { "epoch": 2.92, "learning_rate": 1.4989181825466785e-05, "loss": 0.0006, "step": 20250 }, { "epoch": 2.92, "learning_rate": 1.4979164997195289e-05, "loss": 0.0008, "step": 20255 }, { "epoch": 2.92, "learning_rate": 1.4969148168923794e-05, "loss": 0.0007, "step": 20260 }, { "epoch": 2.92, "learning_rate": 1.4959131340652296e-05, "loss": 0.0006, "step": 20265 }, { "epoch": 2.92, "learning_rate": 1.49491145123808e-05, "loss": 0.0007, "step": 20270 }, { "epoch": 2.92, "learning_rate": 1.4939097684109305e-05, "loss": 0.0008, "step": 20275 }, { "epoch": 2.93, "learning_rate": 1.4929080855837807e-05, "loss": 0.0008, "step": 20280 }, { "epoch": 2.93, "learning_rate": 1.4919064027566312e-05, "loss": 0.0013, "step": 20285 }, { "epoch": 2.93, "learning_rate": 1.4909047199294815e-05, "loss": 0.0008, "step": 20290 }, { "epoch": 2.93, "learning_rate": 1.489903037102332e-05, "loss": 0.0006, "step": 20295 }, { "epoch": 2.93, "learning_rate": 1.4889013542751823e-05, "loss": 0.0006, "step": 20300 }, { "epoch": 2.93, "learning_rate": 1.4878996714480328e-05, "loss": 0.0009, "step": 20305 }, { "epoch": 2.93, "learning_rate": 1.4868979886208831e-05, "loss": 0.0008, "step": 20310 }, { "epoch": 2.93, "learning_rate": 1.4858963057937336e-05, "loss": 0.0008, "step": 20315 }, { "epoch": 2.93, "learning_rate": 1.4848946229665838e-05, "loss": 0.0006, "step": 20320 }, { "epoch": 2.93, "learning_rate": 1.4838929401394344e-05, "loss": 0.001, "step": 20325 }, { "epoch": 2.93, "learning_rate": 1.4828912573122847e-05, "loss": 0.0008, "step": 20330 }, { "epoch": 2.93, "learning_rate": 1.4818895744851352e-05, "loss": 0.0005, "step": 20335 }, { "epoch": 2.93, "learning_rate": 1.4808878916579854e-05, "loss": 0.001, "step": 20340 }, { "epoch": 2.93, "learning_rate": 1.479886208830836e-05, "loss": 0.0008, "step": 20345 }, { "epoch": 2.94, "learning_rate": 1.4788845260036863e-05, "loss": 0.0007, "step": 20350 }, { "epoch": 2.94, "learning_rate": 1.4778828431765365e-05, "loss": 0.0007, "step": 20355 }, { "epoch": 2.94, "learning_rate": 1.476881160349387e-05, "loss": 0.0007, "step": 20360 }, { "epoch": 2.94, "learning_rate": 1.4758794775222374e-05, "loss": 0.0006, "step": 20365 }, { "epoch": 2.94, "learning_rate": 1.4748777946950879e-05, "loss": 0.0011, "step": 20370 }, { "epoch": 2.94, "learning_rate": 1.473876111867938e-05, "loss": 0.001, "step": 20375 }, { "epoch": 2.94, "learning_rate": 1.4728744290407886e-05, "loss": 0.0008, "step": 20380 }, { "epoch": 2.94, "learning_rate": 1.471872746213639e-05, "loss": 0.0009, "step": 20385 }, { "epoch": 2.94, "learning_rate": 1.4708710633864895e-05, "loss": 0.0011, "step": 20390 }, { "epoch": 2.94, "learning_rate": 1.4698693805593396e-05, "loss": 0.0008, "step": 20395 }, { "epoch": 2.94, "learning_rate": 1.4688676977321902e-05, "loss": 0.0008, "step": 20400 }, { "epoch": 2.94, "learning_rate": 1.4678660149050405e-05, "loss": 0.0011, "step": 20405 }, { "epoch": 2.94, "learning_rate": 1.466864332077891e-05, "loss": 0.0011, "step": 20410 }, { "epoch": 2.94, "learning_rate": 1.4658626492507412e-05, "loss": 0.0007, "step": 20415 }, { "epoch": 2.95, "learning_rate": 1.4648609664235917e-05, "loss": 0.0009, "step": 20420 }, { "epoch": 2.95, "learning_rate": 1.4638592835964421e-05, "loss": 0.0007, "step": 20425 }, { "epoch": 2.95, "learning_rate": 1.4628576007692923e-05, "loss": 0.0008, "step": 20430 }, { "epoch": 2.95, "learning_rate": 1.4618559179421428e-05, "loss": 0.0009, "step": 20435 }, { "epoch": 2.95, "learning_rate": 1.4608542351149932e-05, "loss": 0.0012, "step": 20440 }, { "epoch": 2.95, "learning_rate": 1.4598525522878437e-05, "loss": 0.0008, "step": 20445 }, { "epoch": 2.95, "learning_rate": 1.4588508694606939e-05, "loss": 0.0006, "step": 20450 }, { "epoch": 2.95, "learning_rate": 1.4578491866335446e-05, "loss": 0.0006, "step": 20455 }, { "epoch": 2.95, "learning_rate": 1.4568475038063947e-05, "loss": 0.001, "step": 20460 }, { "epoch": 2.95, "learning_rate": 1.4558458209792453e-05, "loss": 0.0007, "step": 20465 }, { "epoch": 2.95, "learning_rate": 1.4548441381520954e-05, "loss": 0.001, "step": 20470 }, { "epoch": 2.95, "learning_rate": 1.4538424553249461e-05, "loss": 0.0003, "step": 20475 }, { "epoch": 2.95, "learning_rate": 1.4528407724977963e-05, "loss": 0.0007, "step": 20480 }, { "epoch": 2.95, "learning_rate": 1.4518390896706468e-05, "loss": 0.0009, "step": 20485 }, { "epoch": 2.96, "learning_rate": 1.4508374068434972e-05, "loss": 0.001, "step": 20490 }, { "epoch": 2.96, "learning_rate": 1.4498357240163477e-05, "loss": 0.0007, "step": 20495 }, { "epoch": 2.96, "learning_rate": 1.4488340411891979e-05, "loss": 0.0009, "step": 20500 }, { "epoch": 2.96, "learning_rate": 1.4478323583620484e-05, "loss": 0.001, "step": 20505 }, { "epoch": 2.96, "learning_rate": 1.4468306755348988e-05, "loss": 0.0007, "step": 20510 }, { "epoch": 2.96, "learning_rate": 1.445828992707749e-05, "loss": 0.0008, "step": 20515 }, { "epoch": 2.96, "learning_rate": 1.4448273098805995e-05, "loss": 0.0011, "step": 20520 }, { "epoch": 2.96, "learning_rate": 1.4438256270534498e-05, "loss": 0.0007, "step": 20525 }, { "epoch": 2.96, "learning_rate": 1.4428239442263004e-05, "loss": 0.0009, "step": 20530 }, { "epoch": 2.96, "learning_rate": 1.4418222613991505e-05, "loss": 0.0007, "step": 20535 }, { "epoch": 2.96, "learning_rate": 1.440820578572001e-05, "loss": 0.0013, "step": 20540 }, { "epoch": 2.96, "learning_rate": 1.4398188957448514e-05, "loss": 0.0005, "step": 20545 }, { "epoch": 2.96, "learning_rate": 1.438817212917702e-05, "loss": 0.0009, "step": 20550 }, { "epoch": 2.96, "learning_rate": 1.4378155300905521e-05, "loss": 0.0007, "step": 20555 }, { "epoch": 2.97, "learning_rate": 1.4368138472634027e-05, "loss": 0.0005, "step": 20560 }, { "epoch": 2.97, "learning_rate": 1.435812164436253e-05, "loss": 0.0006, "step": 20565 }, { "epoch": 2.97, "learning_rate": 1.4348104816091035e-05, "loss": 0.0006, "step": 20570 }, { "epoch": 2.97, "learning_rate": 1.4338087987819537e-05, "loss": 0.0007, "step": 20575 }, { "epoch": 2.97, "learning_rate": 1.4328071159548042e-05, "loss": 0.0007, "step": 20580 }, { "epoch": 2.97, "learning_rate": 1.4318054331276546e-05, "loss": 0.0005, "step": 20585 }, { "epoch": 2.97, "learning_rate": 1.4308037503005048e-05, "loss": 0.0008, "step": 20590 }, { "epoch": 2.97, "learning_rate": 1.4298020674733553e-05, "loss": 0.0011, "step": 20595 }, { "epoch": 2.97, "learning_rate": 1.4288003846462056e-05, "loss": 0.0008, "step": 20600 }, { "epoch": 2.97, "learning_rate": 1.4277987018190562e-05, "loss": 0.0008, "step": 20605 }, { "epoch": 2.97, "learning_rate": 1.4267970189919064e-05, "loss": 0.0007, "step": 20610 }, { "epoch": 2.97, "learning_rate": 1.4257953361647569e-05, "loss": 0.0006, "step": 20615 }, { "epoch": 2.97, "learning_rate": 1.4247936533376072e-05, "loss": 0.0007, "step": 20620 }, { "epoch": 2.97, "learning_rate": 1.4237919705104578e-05, "loss": 0.001, "step": 20625 }, { "epoch": 2.98, "learning_rate": 1.422790287683308e-05, "loss": 0.0006, "step": 20630 }, { "epoch": 2.98, "learning_rate": 1.4217886048561585e-05, "loss": 0.0005, "step": 20635 }, { "epoch": 2.98, "learning_rate": 1.4207869220290088e-05, "loss": 0.0008, "step": 20640 }, { "epoch": 2.98, "learning_rate": 1.4197852392018593e-05, "loss": 0.0006, "step": 20645 }, { "epoch": 2.98, "learning_rate": 1.4187835563747095e-05, "loss": 0.0007, "step": 20650 }, { "epoch": 2.98, "learning_rate": 1.41778187354756e-05, "loss": 0.0005, "step": 20655 }, { "epoch": 2.98, "learning_rate": 1.4167801907204104e-05, "loss": 0.001, "step": 20660 }, { "epoch": 2.98, "learning_rate": 1.4157785078932606e-05, "loss": 0.0011, "step": 20665 }, { "epoch": 2.98, "learning_rate": 1.4147768250661111e-05, "loss": 0.0006, "step": 20670 }, { "epoch": 2.98, "learning_rate": 1.4137751422389615e-05, "loss": 0.0007, "step": 20675 }, { "epoch": 2.98, "learning_rate": 1.412773459411812e-05, "loss": 0.0006, "step": 20680 }, { "epoch": 2.98, "learning_rate": 1.4117717765846622e-05, "loss": 0.0006, "step": 20685 }, { "epoch": 2.98, "learning_rate": 1.4107700937575127e-05, "loss": 0.0008, "step": 20690 }, { "epoch": 2.98, "learning_rate": 1.409768410930363e-05, "loss": 0.0006, "step": 20695 }, { "epoch": 2.99, "learning_rate": 1.4087667281032136e-05, "loss": 0.0006, "step": 20700 }, { "epoch": 2.99, "learning_rate": 1.4077650452760637e-05, "loss": 0.0007, "step": 20705 }, { "epoch": 2.99, "learning_rate": 1.4067633624489143e-05, "loss": 0.0006, "step": 20710 }, { "epoch": 2.99, "learning_rate": 1.4057616796217646e-05, "loss": 0.0008, "step": 20715 }, { "epoch": 2.99, "learning_rate": 1.4047599967946151e-05, "loss": 0.0012, "step": 20720 }, { "epoch": 2.99, "learning_rate": 1.4037583139674653e-05, "loss": 0.0007, "step": 20725 }, { "epoch": 2.99, "learning_rate": 1.4027566311403159e-05, "loss": 0.0008, "step": 20730 }, { "epoch": 2.99, "learning_rate": 1.4017549483131662e-05, "loss": 0.0011, "step": 20735 }, { "epoch": 2.99, "learning_rate": 1.4007532654860167e-05, "loss": 0.0007, "step": 20740 }, { "epoch": 2.99, "learning_rate": 1.3997515826588669e-05, "loss": 0.0007, "step": 20745 }, { "epoch": 2.99, "learning_rate": 1.3987498998317173e-05, "loss": 0.0009, "step": 20750 }, { "epoch": 2.99, "learning_rate": 1.3977482170045678e-05, "loss": 0.0009, "step": 20755 }, { "epoch": 2.99, "learning_rate": 1.396746534177418e-05, "loss": 0.0008, "step": 20760 }, { "epoch": 3.0, "learning_rate": 1.3957448513502685e-05, "loss": 0.0007, "step": 20765 }, { "epoch": 3.0, "learning_rate": 1.3947431685231188e-05, "loss": 0.0007, "step": 20770 }, { "epoch": 3.0, "learning_rate": 1.3937414856959694e-05, "loss": 0.0005, "step": 20775 }, { "epoch": 3.0, "learning_rate": 1.3927398028688196e-05, "loss": 0.0012, "step": 20780 }, { "epoch": 3.0, "learning_rate": 1.39173812004167e-05, "loss": 0.0009, "step": 20785 }, { "epoch": 3.0, "learning_rate": 1.3907364372145204e-05, "loss": 0.001, "step": 20790 }, { "epoch": 3.0, "learning_rate": 1.389734754387371e-05, "loss": 0.0014, "step": 20795 }, { "epoch": 3.0, "learning_rate": 1.3887330715602211e-05, "loss": 0.001, "step": 20800 }, { "epoch": 3.0, "learning_rate": 1.3877313887330717e-05, "loss": 0.0009, "step": 20805 }, { "epoch": 3.0, "learning_rate": 1.386729705905922e-05, "loss": 0.0006, "step": 20810 }, { "epoch": 3.0, "learning_rate": 1.3857280230787725e-05, "loss": 0.0005, "step": 20815 }, { "epoch": 3.0, "learning_rate": 1.3847263402516227e-05, "loss": 0.0006, "step": 20820 }, { "epoch": 3.0, "learning_rate": 1.383724657424473e-05, "loss": 0.0007, "step": 20825 }, { "epoch": 3.0, "learning_rate": 1.3827229745973236e-05, "loss": 0.0008, "step": 20830 }, { "epoch": 3.01, "learning_rate": 1.3817212917701738e-05, "loss": 0.0008, "step": 20835 }, { "epoch": 3.01, "learning_rate": 1.3807196089430243e-05, "loss": 0.0005, "step": 20840 }, { "epoch": 3.01, "learning_rate": 1.3797179261158747e-05, "loss": 0.0008, "step": 20845 }, { "epoch": 3.01, "learning_rate": 1.3787162432887252e-05, "loss": 0.001, "step": 20850 }, { "epoch": 3.01, "learning_rate": 1.3777145604615754e-05, "loss": 0.0007, "step": 20855 }, { "epoch": 3.01, "learning_rate": 1.376712877634426e-05, "loss": 0.0005, "step": 20860 }, { "epoch": 3.01, "learning_rate": 1.3757111948072762e-05, "loss": 0.0004, "step": 20865 }, { "epoch": 3.01, "learning_rate": 1.3747095119801268e-05, "loss": 0.0008, "step": 20870 }, { "epoch": 3.01, "learning_rate": 1.373707829152977e-05, "loss": 0.0007, "step": 20875 }, { "epoch": 3.01, "learning_rate": 1.3727061463258276e-05, "loss": 0.0008, "step": 20880 }, { "epoch": 3.01, "learning_rate": 1.3717044634986778e-05, "loss": 0.0006, "step": 20885 }, { "epoch": 3.01, "learning_rate": 1.3707027806715283e-05, "loss": 0.0006, "step": 20890 }, { "epoch": 3.01, "learning_rate": 1.3697010978443787e-05, "loss": 0.0008, "step": 20895 }, { "epoch": 3.01, "learning_rate": 1.3686994150172292e-05, "loss": 0.0008, "step": 20900 }, { "epoch": 3.02, "learning_rate": 1.3676977321900794e-05, "loss": 0.0006, "step": 20905 }, { "epoch": 3.02, "learning_rate": 1.3666960493629296e-05, "loss": 0.0008, "step": 20910 }, { "epoch": 3.02, "learning_rate": 1.3656943665357803e-05, "loss": 0.0009, "step": 20915 }, { "epoch": 3.02, "learning_rate": 1.3646926837086305e-05, "loss": 0.0008, "step": 20920 }, { "epoch": 3.02, "learning_rate": 1.363691000881481e-05, "loss": 0.0007, "step": 20925 }, { "epoch": 3.02, "learning_rate": 1.3626893180543313e-05, "loss": 0.0005, "step": 20930 }, { "epoch": 3.02, "learning_rate": 1.3616876352271819e-05, "loss": 0.0007, "step": 20935 }, { "epoch": 3.02, "learning_rate": 1.360685952400032e-05, "loss": 0.0008, "step": 20940 }, { "epoch": 3.02, "learning_rate": 1.3596842695728826e-05, "loss": 0.0005, "step": 20945 }, { "epoch": 3.02, "learning_rate": 1.358682586745733e-05, "loss": 0.0007, "step": 20950 }, { "epoch": 3.02, "learning_rate": 1.3576809039185834e-05, "loss": 0.0007, "step": 20955 }, { "epoch": 3.02, "learning_rate": 1.3566792210914336e-05, "loss": 0.0006, "step": 20960 }, { "epoch": 3.02, "learning_rate": 1.3556775382642841e-05, "loss": 0.0008, "step": 20965 }, { "epoch": 3.02, "learning_rate": 1.3546758554371345e-05, "loss": 0.001, "step": 20970 }, { "epoch": 3.03, "learning_rate": 1.353674172609985e-05, "loss": 0.0008, "step": 20975 }, { "epoch": 3.03, "learning_rate": 1.3526724897828352e-05, "loss": 0.001, "step": 20980 }, { "epoch": 3.03, "learning_rate": 1.3516708069556856e-05, "loss": 0.0007, "step": 20985 }, { "epoch": 3.03, "learning_rate": 1.350669124128536e-05, "loss": 0.0008, "step": 20990 }, { "epoch": 3.03, "learning_rate": 1.3496674413013863e-05, "loss": 0.0005, "step": 20995 }, { "epoch": 3.03, "learning_rate": 1.3486657584742368e-05, "loss": 0.001, "step": 21000 }, { "epoch": 3.03, "learning_rate": 1.3476640756470871e-05, "loss": 0.0007, "step": 21005 }, { "epoch": 3.03, "learning_rate": 1.3466623928199377e-05, "loss": 0.0009, "step": 21010 }, { "epoch": 3.03, "learning_rate": 1.3456607099927878e-05, "loss": 0.0006, "step": 21015 }, { "epoch": 3.03, "learning_rate": 1.3446590271656384e-05, "loss": 0.0007, "step": 21020 }, { "epoch": 3.03, "learning_rate": 1.3436573443384887e-05, "loss": 0.0007, "step": 21025 }, { "epoch": 3.03, "learning_rate": 1.3426556615113392e-05, "loss": 0.0008, "step": 21030 }, { "epoch": 3.03, "learning_rate": 1.3416539786841894e-05, "loss": 0.0006, "step": 21035 }, { "epoch": 3.03, "learning_rate": 1.34065229585704e-05, "loss": 0.0008, "step": 21040 }, { "epoch": 3.04, "learning_rate": 1.3396506130298903e-05, "loss": 0.0009, "step": 21045 }, { "epoch": 3.04, "learning_rate": 1.3386489302027408e-05, "loss": 0.0006, "step": 21050 }, { "epoch": 3.04, "learning_rate": 1.337647247375591e-05, "loss": 0.0009, "step": 21055 }, { "epoch": 3.04, "learning_rate": 1.3366455645484414e-05, "loss": 0.0005, "step": 21060 }, { "epoch": 3.04, "learning_rate": 1.3356438817212919e-05, "loss": 0.0005, "step": 21065 }, { "epoch": 3.04, "learning_rate": 1.334642198894142e-05, "loss": 0.0009, "step": 21070 }, { "epoch": 3.04, "learning_rate": 1.3336405160669926e-05, "loss": 0.0007, "step": 21075 }, { "epoch": 3.04, "learning_rate": 1.332638833239843e-05, "loss": 0.0006, "step": 21080 }, { "epoch": 3.04, "learning_rate": 1.3316371504126935e-05, "loss": 0.0008, "step": 21085 }, { "epoch": 3.04, "learning_rate": 1.3306354675855437e-05, "loss": 0.0009, "step": 21090 }, { "epoch": 3.04, "learning_rate": 1.3296337847583942e-05, "loss": 0.0005, "step": 21095 }, { "epoch": 3.04, "learning_rate": 1.3286321019312445e-05, "loss": 0.0004, "step": 21100 }, { "epoch": 3.04, "learning_rate": 1.327630419104095e-05, "loss": 0.001, "step": 21105 }, { "epoch": 3.04, "learning_rate": 1.3266287362769452e-05, "loss": 0.0005, "step": 21110 }, { "epoch": 3.05, "learning_rate": 1.3256270534497958e-05, "loss": 0.0007, "step": 21115 }, { "epoch": 3.05, "learning_rate": 1.3246253706226461e-05, "loss": 0.001, "step": 21120 }, { "epoch": 3.05, "learning_rate": 1.3236236877954966e-05, "loss": 0.0008, "step": 21125 }, { "epoch": 3.05, "learning_rate": 1.3226220049683468e-05, "loss": 0.0007, "step": 21130 }, { "epoch": 3.05, "learning_rate": 1.3216203221411973e-05, "loss": 0.0007, "step": 21135 }, { "epoch": 3.05, "learning_rate": 1.3206186393140477e-05, "loss": 0.0006, "step": 21140 }, { "epoch": 3.05, "learning_rate": 1.3196169564868979e-05, "loss": 0.0012, "step": 21145 }, { "epoch": 3.05, "learning_rate": 1.3186152736597484e-05, "loss": 0.001, "step": 21150 }, { "epoch": 3.05, "learning_rate": 1.3176135908325988e-05, "loss": 0.0008, "step": 21155 }, { "epoch": 3.05, "learning_rate": 1.3166119080054493e-05, "loss": 0.0008, "step": 21160 }, { "epoch": 3.05, "learning_rate": 1.3156102251782995e-05, "loss": 0.0007, "step": 21165 }, { "epoch": 3.05, "learning_rate": 1.31460854235115e-05, "loss": 0.0007, "step": 21170 }, { "epoch": 3.05, "learning_rate": 1.3136068595240003e-05, "loss": 0.0008, "step": 21175 }, { "epoch": 3.05, "learning_rate": 1.3126051766968509e-05, "loss": 0.0003, "step": 21180 }, { "epoch": 3.06, "learning_rate": 1.311603493869701e-05, "loss": 0.0006, "step": 21185 }, { "epoch": 3.06, "learning_rate": 1.3106018110425516e-05, "loss": 0.001, "step": 21190 }, { "epoch": 3.06, "learning_rate": 1.309600128215402e-05, "loss": 0.001, "step": 21195 }, { "epoch": 3.06, "learning_rate": 1.3085984453882524e-05, "loss": 0.0005, "step": 21200 }, { "epoch": 3.06, "learning_rate": 1.3075967625611026e-05, "loss": 0.0009, "step": 21205 }, { "epoch": 3.06, "learning_rate": 1.3065950797339532e-05, "loss": 0.0007, "step": 21210 }, { "epoch": 3.06, "learning_rate": 1.3055933969068035e-05, "loss": 0.0008, "step": 21215 }, { "epoch": 3.06, "learning_rate": 1.3045917140796537e-05, "loss": 0.0006, "step": 21220 }, { "epoch": 3.06, "learning_rate": 1.3035900312525042e-05, "loss": 0.0005, "step": 21225 }, { "epoch": 3.06, "learning_rate": 1.3025883484253546e-05, "loss": 0.0008, "step": 21230 }, { "epoch": 3.06, "learning_rate": 1.3015866655982051e-05, "loss": 0.001, "step": 21235 }, { "epoch": 3.06, "learning_rate": 1.3005849827710553e-05, "loss": 0.0007, "step": 21240 }, { "epoch": 3.06, "learning_rate": 1.2995832999439058e-05, "loss": 0.0004, "step": 21245 }, { "epoch": 3.07, "learning_rate": 1.2985816171167561e-05, "loss": 0.0007, "step": 21250 }, { "epoch": 3.07, "learning_rate": 1.2975799342896067e-05, "loss": 0.0006, "step": 21255 }, { "epoch": 3.07, "learning_rate": 1.2965782514624569e-05, "loss": 0.0008, "step": 21260 }, { "epoch": 3.07, "learning_rate": 1.2955765686353075e-05, "loss": 0.0008, "step": 21265 }, { "epoch": 3.07, "learning_rate": 1.2945748858081577e-05, "loss": 0.0007, "step": 21270 }, { "epoch": 3.07, "learning_rate": 1.2935732029810083e-05, "loss": 0.0006, "step": 21275 }, { "epoch": 3.07, "learning_rate": 1.2925715201538584e-05, "loss": 0.0007, "step": 21280 }, { "epoch": 3.07, "learning_rate": 1.2915698373267091e-05, "loss": 0.0007, "step": 21285 }, { "epoch": 3.07, "learning_rate": 1.2905681544995593e-05, "loss": 0.0005, "step": 21290 }, { "epoch": 3.07, "learning_rate": 1.2895664716724098e-05, "loss": 0.0004, "step": 21295 }, { "epoch": 3.07, "learning_rate": 1.2885647888452602e-05, "loss": 0.0005, "step": 21300 }, { "epoch": 3.07, "learning_rate": 1.2875631060181104e-05, "loss": 0.0004, "step": 21305 }, { "epoch": 3.07, "learning_rate": 1.2865614231909609e-05, "loss": 0.0008, "step": 21310 }, { "epoch": 3.07, "learning_rate": 1.285559740363811e-05, "loss": 0.0007, "step": 21315 }, { "epoch": 3.08, "learning_rate": 1.2845580575366618e-05, "loss": 0.0009, "step": 21320 }, { "epoch": 3.08, "learning_rate": 1.283556374709512e-05, "loss": 0.0009, "step": 21325 }, { "epoch": 3.08, "learning_rate": 1.2825546918823625e-05, "loss": 0.001, "step": 21330 }, { "epoch": 3.08, "learning_rate": 1.2815530090552128e-05, "loss": 0.0007, "step": 21335 }, { "epoch": 3.08, "learning_rate": 1.2805513262280634e-05, "loss": 0.0008, "step": 21340 }, { "epoch": 3.08, "learning_rate": 1.2795496434009135e-05, "loss": 0.001, "step": 21345 }, { "epoch": 3.08, "learning_rate": 1.278547960573764e-05, "loss": 0.0008, "step": 21350 }, { "epoch": 3.08, "learning_rate": 1.2775462777466144e-05, "loss": 0.0006, "step": 21355 }, { "epoch": 3.08, "learning_rate": 1.276544594919465e-05, "loss": 0.0006, "step": 21360 }, { "epoch": 3.08, "learning_rate": 1.2755429120923151e-05, "loss": 0.0007, "step": 21365 }, { "epoch": 3.08, "learning_rate": 1.2745412292651656e-05, "loss": 0.001, "step": 21370 }, { "epoch": 3.08, "learning_rate": 1.273539546438016e-05, "loss": 0.0008, "step": 21375 }, { "epoch": 3.08, "learning_rate": 1.2725378636108662e-05, "loss": 0.0005, "step": 21380 }, { "epoch": 3.08, "learning_rate": 1.2715361807837167e-05, "loss": 0.0009, "step": 21385 }, { "epoch": 3.09, "learning_rate": 1.270534497956567e-05, "loss": 0.0004, "step": 21390 }, { "epoch": 3.09, "learning_rate": 1.2695328151294176e-05, "loss": 0.0011, "step": 21395 }, { "epoch": 3.09, "learning_rate": 1.2685311323022678e-05, "loss": 0.0006, "step": 21400 }, { "epoch": 3.09, "learning_rate": 1.2675294494751183e-05, "loss": 0.0006, "step": 21405 }, { "epoch": 3.09, "learning_rate": 1.2665277666479686e-05, "loss": 0.0008, "step": 21410 }, { "epoch": 3.09, "learning_rate": 1.2655260838208192e-05, "loss": 0.0007, "step": 21415 }, { "epoch": 3.09, "learning_rate": 1.2645244009936693e-05, "loss": 0.0011, "step": 21420 }, { "epoch": 3.09, "learning_rate": 1.2635227181665199e-05, "loss": 0.0004, "step": 21425 }, { "epoch": 3.09, "learning_rate": 1.2625210353393702e-05, "loss": 0.0005, "step": 21430 }, { "epoch": 3.09, "learning_rate": 1.2615193525122207e-05, "loss": 0.0009, "step": 21435 }, { "epoch": 3.09, "learning_rate": 1.260517669685071e-05, "loss": 0.0011, "step": 21440 }, { "epoch": 3.09, "learning_rate": 1.2595159868579214e-05, "loss": 0.0008, "step": 21445 }, { "epoch": 3.09, "learning_rate": 1.2585143040307718e-05, "loss": 0.0006, "step": 21450 }, { "epoch": 3.09, "learning_rate": 1.257512621203622e-05, "loss": 0.0006, "step": 21455 }, { "epoch": 3.1, "learning_rate": 1.2565109383764725e-05, "loss": 0.0006, "step": 21460 }, { "epoch": 3.1, "learning_rate": 1.2555092555493229e-05, "loss": 0.0007, "step": 21465 }, { "epoch": 3.1, "learning_rate": 1.2545075727221734e-05, "loss": 0.0004, "step": 21470 }, { "epoch": 3.1, "learning_rate": 1.2535058898950236e-05, "loss": 0.0007, "step": 21475 }, { "epoch": 3.1, "learning_rate": 1.2525042070678741e-05, "loss": 0.0006, "step": 21480 }, { "epoch": 3.1, "learning_rate": 1.2515025242407244e-05, "loss": 0.0007, "step": 21485 }, { "epoch": 3.1, "learning_rate": 1.250500841413575e-05, "loss": 0.001, "step": 21490 }, { "epoch": 3.1, "learning_rate": 1.2494991585864252e-05, "loss": 0.0004, "step": 21495 }, { "epoch": 3.1, "learning_rate": 1.2484974757592755e-05, "loss": 0.0007, "step": 21500 }, { "epoch": 3.1, "learning_rate": 1.247495792932126e-05, "loss": 0.0009, "step": 21505 }, { "epoch": 3.1, "learning_rate": 1.2464941101049764e-05, "loss": 0.0009, "step": 21510 }, { "epoch": 3.1, "learning_rate": 1.2454924272778267e-05, "loss": 0.0005, "step": 21515 }, { "epoch": 3.1, "learning_rate": 1.2444907444506773e-05, "loss": 0.0008, "step": 21520 }, { "epoch": 3.1, "learning_rate": 1.2434890616235276e-05, "loss": 0.0004, "step": 21525 }, { "epoch": 3.11, "learning_rate": 1.242487378796378e-05, "loss": 0.0009, "step": 21530 }, { "epoch": 3.11, "learning_rate": 1.2414856959692283e-05, "loss": 0.0008, "step": 21535 }, { "epoch": 3.11, "learning_rate": 1.2404840131420788e-05, "loss": 0.0009, "step": 21540 }, { "epoch": 3.11, "learning_rate": 1.2394823303149292e-05, "loss": 0.0008, "step": 21545 }, { "epoch": 3.11, "learning_rate": 1.2384806474877795e-05, "loss": 0.0008, "step": 21550 }, { "epoch": 3.11, "learning_rate": 1.2374789646606299e-05, "loss": 0.0006, "step": 21555 }, { "epoch": 3.11, "learning_rate": 1.2364772818334804e-05, "loss": 0.0007, "step": 21560 }, { "epoch": 3.11, "learning_rate": 1.2354755990063308e-05, "loss": 0.0009, "step": 21565 }, { "epoch": 3.11, "learning_rate": 1.2344739161791811e-05, "loss": 0.0003, "step": 21570 }, { "epoch": 3.11, "learning_rate": 1.2334722333520315e-05, "loss": 0.0005, "step": 21575 }, { "epoch": 3.11, "learning_rate": 1.2324705505248818e-05, "loss": 0.0009, "step": 21580 }, { "epoch": 3.11, "learning_rate": 1.2314688676977322e-05, "loss": 0.0011, "step": 21585 }, { "epoch": 3.11, "learning_rate": 1.2304671848705825e-05, "loss": 0.0004, "step": 21590 }, { "epoch": 3.11, "learning_rate": 1.229465502043433e-05, "loss": 0.0007, "step": 21595 }, { "epoch": 3.12, "learning_rate": 1.2284638192162834e-05, "loss": 0.0008, "step": 21600 }, { "epoch": 3.12, "learning_rate": 1.2274621363891338e-05, "loss": 0.0007, "step": 21605 }, { "epoch": 3.12, "learning_rate": 1.2264604535619841e-05, "loss": 0.0006, "step": 21610 }, { "epoch": 3.12, "learning_rate": 1.2254587707348346e-05, "loss": 0.0008, "step": 21615 }, { "epoch": 3.12, "learning_rate": 1.224457087907685e-05, "loss": 0.0008, "step": 21620 }, { "epoch": 3.12, "learning_rate": 1.2234554050805354e-05, "loss": 0.0005, "step": 21625 }, { "epoch": 3.12, "learning_rate": 1.2224537222533857e-05, "loss": 0.0011, "step": 21630 }, { "epoch": 3.12, "learning_rate": 1.2214520394262362e-05, "loss": 0.0009, "step": 21635 }, { "epoch": 3.12, "learning_rate": 1.2204503565990866e-05, "loss": 0.0008, "step": 21640 }, { "epoch": 3.12, "learning_rate": 1.219448673771937e-05, "loss": 0.0007, "step": 21645 }, { "epoch": 3.12, "learning_rate": 1.2184469909447873e-05, "loss": 0.0009, "step": 21650 }, { "epoch": 3.12, "learning_rate": 1.2174453081176376e-05, "loss": 0.0007, "step": 21655 }, { "epoch": 3.12, "learning_rate": 1.216443625290488e-05, "loss": 0.0005, "step": 21660 }, { "epoch": 3.12, "learning_rate": 1.2154419424633383e-05, "loss": 0.0007, "step": 21665 }, { "epoch": 3.13, "learning_rate": 1.2144402596361889e-05, "loss": 0.001, "step": 21670 }, { "epoch": 3.13, "learning_rate": 1.2134385768090392e-05, "loss": 0.001, "step": 21675 }, { "epoch": 3.13, "learning_rate": 1.2124368939818896e-05, "loss": 0.0008, "step": 21680 }, { "epoch": 3.13, "learning_rate": 1.21143521115474e-05, "loss": 0.0009, "step": 21685 }, { "epoch": 3.13, "learning_rate": 1.2104335283275905e-05, "loss": 0.0006, "step": 21690 }, { "epoch": 3.13, "learning_rate": 1.2094318455004408e-05, "loss": 0.0005, "step": 21695 }, { "epoch": 3.13, "learning_rate": 1.2084301626732912e-05, "loss": 0.0009, "step": 21700 }, { "epoch": 3.13, "learning_rate": 1.2074284798461417e-05, "loss": 0.0004, "step": 21705 }, { "epoch": 3.13, "learning_rate": 1.206426797018992e-05, "loss": 0.0009, "step": 21710 }, { "epoch": 3.13, "learning_rate": 1.2054251141918424e-05, "loss": 0.001, "step": 21715 }, { "epoch": 3.13, "learning_rate": 1.2044234313646927e-05, "loss": 0.0011, "step": 21720 }, { "epoch": 3.13, "learning_rate": 1.2034217485375433e-05, "loss": 0.0011, "step": 21725 }, { "epoch": 3.13, "learning_rate": 1.2024200657103934e-05, "loss": 0.0008, "step": 21730 }, { "epoch": 3.14, "learning_rate": 1.2014183828832438e-05, "loss": 0.0007, "step": 21735 }, { "epoch": 3.14, "learning_rate": 1.2004167000560943e-05, "loss": 0.0008, "step": 21740 }, { "epoch": 3.14, "learning_rate": 1.1994150172289447e-05, "loss": 0.0005, "step": 21745 }, { "epoch": 3.14, "learning_rate": 1.198413334401795e-05, "loss": 0.0008, "step": 21750 }, { "epoch": 3.14, "learning_rate": 1.1974116515746454e-05, "loss": 0.0009, "step": 21755 }, { "epoch": 3.14, "learning_rate": 1.1964099687474959e-05, "loss": 0.0004, "step": 21760 }, { "epoch": 3.14, "learning_rate": 1.1954082859203463e-05, "loss": 0.0007, "step": 21765 }, { "epoch": 3.14, "learning_rate": 1.1944066030931966e-05, "loss": 0.0006, "step": 21770 }, { "epoch": 3.14, "learning_rate": 1.193404920266047e-05, "loss": 0.0009, "step": 21775 }, { "epoch": 3.14, "learning_rate": 1.1924032374388975e-05, "loss": 0.0007, "step": 21780 }, { "epoch": 3.14, "learning_rate": 1.1914015546117478e-05, "loss": 0.0006, "step": 21785 }, { "epoch": 3.14, "learning_rate": 1.1903998717845982e-05, "loss": 0.0007, "step": 21790 }, { "epoch": 3.14, "learning_rate": 1.1893981889574486e-05, "loss": 0.0007, "step": 21795 }, { "epoch": 3.14, "learning_rate": 1.188396506130299e-05, "loss": 0.0009, "step": 21800 }, { "epoch": 3.15, "learning_rate": 1.1873948233031494e-05, "loss": 0.0007, "step": 21805 }, { "epoch": 3.15, "learning_rate": 1.1863931404759996e-05, "loss": 0.0005, "step": 21810 }, { "epoch": 3.15, "learning_rate": 1.1853914576488501e-05, "loss": 0.0011, "step": 21815 }, { "epoch": 3.15, "learning_rate": 1.1843897748217005e-05, "loss": 0.0007, "step": 21820 }, { "epoch": 3.15, "learning_rate": 1.1833880919945508e-05, "loss": 0.0009, "step": 21825 }, { "epoch": 3.15, "learning_rate": 1.1823864091674012e-05, "loss": 0.0006, "step": 21830 }, { "epoch": 3.15, "learning_rate": 1.1813847263402517e-05, "loss": 0.0009, "step": 21835 }, { "epoch": 3.15, "learning_rate": 1.180383043513102e-05, "loss": 0.0008, "step": 21840 }, { "epoch": 3.15, "learning_rate": 1.1793813606859524e-05, "loss": 0.0008, "step": 21845 }, { "epoch": 3.15, "learning_rate": 1.1783796778588028e-05, "loss": 0.0008, "step": 21850 }, { "epoch": 3.15, "learning_rate": 1.1773779950316533e-05, "loss": 0.0007, "step": 21855 }, { "epoch": 3.15, "learning_rate": 1.1763763122045037e-05, "loss": 0.0006, "step": 21860 }, { "epoch": 3.15, "learning_rate": 1.175374629377354e-05, "loss": 0.0005, "step": 21865 }, { "epoch": 3.15, "learning_rate": 1.1743729465502044e-05, "loss": 0.0009, "step": 21870 }, { "epoch": 3.16, "learning_rate": 1.1733712637230549e-05, "loss": 0.001, "step": 21875 }, { "epoch": 3.16, "learning_rate": 1.1723695808959052e-05, "loss": 0.0008, "step": 21880 }, { "epoch": 3.16, "learning_rate": 1.1713678980687556e-05, "loss": 0.001, "step": 21885 }, { "epoch": 3.16, "learning_rate": 1.170366215241606e-05, "loss": 0.0009, "step": 21890 }, { "epoch": 3.16, "learning_rate": 1.1693645324144563e-05, "loss": 0.0006, "step": 21895 }, { "epoch": 3.16, "learning_rate": 1.1683628495873066e-05, "loss": 0.0007, "step": 21900 }, { "epoch": 3.16, "learning_rate": 1.167361166760157e-05, "loss": 0.0005, "step": 21905 }, { "epoch": 3.16, "learning_rate": 1.1663594839330075e-05, "loss": 0.0004, "step": 21910 }, { "epoch": 3.16, "learning_rate": 1.1653578011058579e-05, "loss": 0.0004, "step": 21915 }, { "epoch": 3.16, "learning_rate": 1.1643561182787082e-05, "loss": 0.0005, "step": 21920 }, { "epoch": 3.16, "learning_rate": 1.1633544354515588e-05, "loss": 0.0006, "step": 21925 }, { "epoch": 3.16, "learning_rate": 1.1623527526244091e-05, "loss": 0.0012, "step": 21930 }, { "epoch": 3.16, "learning_rate": 1.1613510697972595e-05, "loss": 0.0007, "step": 21935 }, { "epoch": 3.16, "learning_rate": 1.1603493869701098e-05, "loss": 0.0007, "step": 21940 }, { "epoch": 3.17, "learning_rate": 1.1593477041429603e-05, "loss": 0.0007, "step": 21945 }, { "epoch": 3.17, "learning_rate": 1.1583460213158107e-05, "loss": 0.0006, "step": 21950 }, { "epoch": 3.17, "learning_rate": 1.157344338488661e-05, "loss": 0.0007, "step": 21955 }, { "epoch": 3.17, "learning_rate": 1.1563426556615114e-05, "loss": 0.0005, "step": 21960 }, { "epoch": 3.17, "learning_rate": 1.155340972834362e-05, "loss": 0.0009, "step": 21965 }, { "epoch": 3.17, "learning_rate": 1.1543392900072121e-05, "loss": 0.0007, "step": 21970 }, { "epoch": 3.17, "learning_rate": 1.1533376071800625e-05, "loss": 0.0007, "step": 21975 }, { "epoch": 3.17, "learning_rate": 1.152335924352913e-05, "loss": 0.0008, "step": 21980 }, { "epoch": 3.17, "learning_rate": 1.1513342415257633e-05, "loss": 0.0008, "step": 21985 }, { "epoch": 3.17, "learning_rate": 1.1503325586986137e-05, "loss": 0.0007, "step": 21990 }, { "epoch": 3.17, "learning_rate": 1.149330875871464e-05, "loss": 0.0006, "step": 21995 }, { "epoch": 3.17, "learning_rate": 1.1483291930443146e-05, "loss": 0.0009, "step": 22000 }, { "epoch": 3.17, "learning_rate": 1.1473275102171649e-05, "loss": 0.001, "step": 22005 }, { "epoch": 3.17, "learning_rate": 1.1463258273900153e-05, "loss": 0.0009, "step": 22010 }, { "epoch": 3.18, "learning_rate": 1.1453241445628656e-05, "loss": 0.0007, "step": 22015 }, { "epoch": 3.18, "learning_rate": 1.1443224617357161e-05, "loss": 0.0007, "step": 22020 }, { "epoch": 3.18, "learning_rate": 1.1433207789085665e-05, "loss": 0.0008, "step": 22025 }, { "epoch": 3.18, "learning_rate": 1.1423190960814168e-05, "loss": 0.0004, "step": 22030 }, { "epoch": 3.18, "learning_rate": 1.1413174132542672e-05, "loss": 0.0009, "step": 22035 }, { "epoch": 3.18, "learning_rate": 1.1403157304271177e-05, "loss": 0.0006, "step": 22040 }, { "epoch": 3.18, "learning_rate": 1.1393140475999679e-05, "loss": 0.0007, "step": 22045 }, { "epoch": 3.18, "learning_rate": 1.1383123647728183e-05, "loss": 0.0005, "step": 22050 }, { "epoch": 3.18, "learning_rate": 1.1373106819456688e-05, "loss": 0.0008, "step": 22055 }, { "epoch": 3.18, "learning_rate": 1.1363089991185191e-05, "loss": 0.0005, "step": 22060 }, { "epoch": 3.18, "learning_rate": 1.1353073162913695e-05, "loss": 0.0006, "step": 22065 }, { "epoch": 3.18, "learning_rate": 1.1343056334642198e-05, "loss": 0.0008, "step": 22070 }, { "epoch": 3.18, "learning_rate": 1.1333039506370704e-05, "loss": 0.0007, "step": 22075 }, { "epoch": 3.18, "learning_rate": 1.1323022678099207e-05, "loss": 0.0009, "step": 22080 }, { "epoch": 3.19, "learning_rate": 1.131300584982771e-05, "loss": 0.0009, "step": 22085 }, { "epoch": 3.19, "learning_rate": 1.1302989021556214e-05, "loss": 0.0008, "step": 22090 }, { "epoch": 3.19, "learning_rate": 1.129297219328472e-05, "loss": 0.0009, "step": 22095 }, { "epoch": 3.19, "learning_rate": 1.1282955365013223e-05, "loss": 0.0006, "step": 22100 }, { "epoch": 3.19, "learning_rate": 1.1272938536741727e-05, "loss": 0.001, "step": 22105 }, { "epoch": 3.19, "learning_rate": 1.1262921708470232e-05, "loss": 0.0006, "step": 22110 }, { "epoch": 3.19, "learning_rate": 1.1252904880198735e-05, "loss": 0.0009, "step": 22115 }, { "epoch": 3.19, "learning_rate": 1.1242888051927239e-05, "loss": 0.0007, "step": 22120 }, { "epoch": 3.19, "learning_rate": 1.123287122365574e-05, "loss": 0.0006, "step": 22125 }, { "epoch": 3.19, "learning_rate": 1.1222854395384246e-05, "loss": 0.0007, "step": 22130 }, { "epoch": 3.19, "learning_rate": 1.121283756711275e-05, "loss": 0.0009, "step": 22135 }, { "epoch": 3.19, "learning_rate": 1.1202820738841253e-05, "loss": 0.0004, "step": 22140 }, { "epoch": 3.19, "learning_rate": 1.1192803910569758e-05, "loss": 0.001, "step": 22145 }, { "epoch": 3.19, "learning_rate": 1.1182787082298262e-05, "loss": 0.0011, "step": 22150 }, { "epoch": 3.2, "learning_rate": 1.1172770254026765e-05, "loss": 0.001, "step": 22155 }, { "epoch": 3.2, "learning_rate": 1.1162753425755269e-05, "loss": 0.0006, "step": 22160 }, { "epoch": 3.2, "learning_rate": 1.1152736597483774e-05, "loss": 0.0007, "step": 22165 }, { "epoch": 3.2, "learning_rate": 1.1142719769212278e-05, "loss": 0.0007, "step": 22170 }, { "epoch": 3.2, "learning_rate": 1.1132702940940781e-05, "loss": 0.0009, "step": 22175 }, { "epoch": 3.2, "learning_rate": 1.1122686112669285e-05, "loss": 0.0005, "step": 22180 }, { "epoch": 3.2, "learning_rate": 1.111266928439779e-05, "loss": 0.0009, "step": 22185 }, { "epoch": 3.2, "learning_rate": 1.1102652456126293e-05, "loss": 0.0012, "step": 22190 }, { "epoch": 3.2, "learning_rate": 1.1092635627854797e-05, "loss": 0.0007, "step": 22195 }, { "epoch": 3.2, "learning_rate": 1.10826187995833e-05, "loss": 0.0005, "step": 22200 }, { "epoch": 3.2, "learning_rate": 1.1072601971311804e-05, "loss": 0.0007, "step": 22205 }, { "epoch": 3.2, "learning_rate": 1.1062585143040308e-05, "loss": 0.0007, "step": 22210 }, { "epoch": 3.2, "learning_rate": 1.1052568314768811e-05, "loss": 0.0007, "step": 22215 }, { "epoch": 3.2, "learning_rate": 1.1042551486497316e-05, "loss": 0.0007, "step": 22220 }, { "epoch": 3.21, "learning_rate": 1.103253465822582e-05, "loss": 0.0007, "step": 22225 }, { "epoch": 3.21, "learning_rate": 1.1022517829954323e-05, "loss": 0.0006, "step": 22230 }, { "epoch": 3.21, "learning_rate": 1.1012501001682827e-05, "loss": 0.0006, "step": 22235 }, { "epoch": 3.21, "learning_rate": 1.1002484173411332e-05, "loss": 0.0007, "step": 22240 }, { "epoch": 3.21, "learning_rate": 1.0992467345139836e-05, "loss": 0.0007, "step": 22245 }, { "epoch": 3.21, "learning_rate": 1.098245051686834e-05, "loss": 0.0007, "step": 22250 }, { "epoch": 3.21, "learning_rate": 1.0972433688596843e-05, "loss": 0.0005, "step": 22255 }, { "epoch": 3.21, "learning_rate": 1.0962416860325348e-05, "loss": 0.0007, "step": 22260 }, { "epoch": 3.21, "learning_rate": 1.0952400032053851e-05, "loss": 0.0011, "step": 22265 }, { "epoch": 3.21, "learning_rate": 1.0942383203782355e-05, "loss": 0.0011, "step": 22270 }, { "epoch": 3.21, "learning_rate": 1.0932366375510859e-05, "loss": 0.0009, "step": 22275 }, { "epoch": 3.21, "learning_rate": 1.0922349547239364e-05, "loss": 0.0009, "step": 22280 }, { "epoch": 3.21, "learning_rate": 1.0912332718967866e-05, "loss": 0.0008, "step": 22285 }, { "epoch": 3.22, "learning_rate": 1.0902315890696369e-05, "loss": 0.0008, "step": 22290 }, { "epoch": 3.22, "learning_rate": 1.0892299062424874e-05, "loss": 0.0008, "step": 22295 }, { "epoch": 3.22, "learning_rate": 1.0882282234153378e-05, "loss": 0.0003, "step": 22300 }, { "epoch": 3.22, "learning_rate": 1.0872265405881881e-05, "loss": 0.0007, "step": 22305 }, { "epoch": 3.22, "learning_rate": 1.0862248577610385e-05, "loss": 0.0008, "step": 22310 }, { "epoch": 3.22, "learning_rate": 1.085223174933889e-05, "loss": 0.001, "step": 22315 }, { "epoch": 3.22, "learning_rate": 1.0842214921067394e-05, "loss": 0.0011, "step": 22320 }, { "epoch": 3.22, "learning_rate": 1.0832198092795897e-05, "loss": 0.0006, "step": 22325 }, { "epoch": 3.22, "learning_rate": 1.0822181264524402e-05, "loss": 0.0008, "step": 22330 }, { "epoch": 3.22, "learning_rate": 1.0812164436252906e-05, "loss": 0.0007, "step": 22335 }, { "epoch": 3.22, "learning_rate": 1.080214760798141e-05, "loss": 0.0008, "step": 22340 }, { "epoch": 3.22, "learning_rate": 1.0792130779709913e-05, "loss": 0.0008, "step": 22345 }, { "epoch": 3.22, "learning_rate": 1.0782113951438418e-05, "loss": 0.0007, "step": 22350 }, { "epoch": 3.22, "learning_rate": 1.0772097123166922e-05, "loss": 0.0009, "step": 22355 }, { "epoch": 3.23, "learning_rate": 1.0762080294895425e-05, "loss": 0.001, "step": 22360 }, { "epoch": 3.23, "learning_rate": 1.0752063466623929e-05, "loss": 0.0008, "step": 22365 }, { "epoch": 3.23, "learning_rate": 1.0742046638352432e-05, "loss": 0.0007, "step": 22370 }, { "epoch": 3.23, "learning_rate": 1.0732029810080936e-05, "loss": 0.0008, "step": 22375 }, { "epoch": 3.23, "learning_rate": 1.072201298180944e-05, "loss": 0.0007, "step": 22380 }, { "epoch": 3.23, "learning_rate": 1.0711996153537945e-05, "loss": 0.0008, "step": 22385 }, { "epoch": 3.23, "learning_rate": 1.0701979325266448e-05, "loss": 0.0006, "step": 22390 }, { "epoch": 3.23, "learning_rate": 1.0691962496994952e-05, "loss": 0.0009, "step": 22395 }, { "epoch": 3.23, "learning_rate": 1.0681945668723455e-05, "loss": 0.0008, "step": 22400 }, { "epoch": 3.23, "learning_rate": 1.067192884045196e-05, "loss": 0.001, "step": 22405 }, { "epoch": 3.23, "learning_rate": 1.0661912012180464e-05, "loss": 0.0008, "step": 22410 }, { "epoch": 3.23, "learning_rate": 1.0651895183908968e-05, "loss": 0.0009, "step": 22415 }, { "epoch": 3.23, "learning_rate": 1.0641878355637471e-05, "loss": 0.0008, "step": 22420 }, { "epoch": 3.23, "learning_rate": 1.0631861527365976e-05, "loss": 0.0009, "step": 22425 }, { "epoch": 3.24, "learning_rate": 1.062184469909448e-05, "loss": 0.001, "step": 22430 }, { "epoch": 3.24, "learning_rate": 1.0611827870822983e-05, "loss": 0.0013, "step": 22435 }, { "epoch": 3.24, "learning_rate": 1.0601811042551487e-05, "loss": 0.0008, "step": 22440 }, { "epoch": 3.24, "learning_rate": 1.059179421427999e-05, "loss": 0.0007, "step": 22445 }, { "epoch": 3.24, "learning_rate": 1.0581777386008494e-05, "loss": 0.0012, "step": 22450 }, { "epoch": 3.24, "learning_rate": 1.0571760557736998e-05, "loss": 0.0007, "step": 22455 }, { "epoch": 3.24, "learning_rate": 1.0561743729465503e-05, "loss": 0.0006, "step": 22460 }, { "epoch": 3.24, "learning_rate": 1.0551726901194006e-05, "loss": 0.0006, "step": 22465 }, { "epoch": 3.24, "learning_rate": 1.054171007292251e-05, "loss": 0.0007, "step": 22470 }, { "epoch": 3.24, "learning_rate": 1.0531693244651013e-05, "loss": 0.0007, "step": 22475 }, { "epoch": 3.24, "learning_rate": 1.0521676416379519e-05, "loss": 0.0006, "step": 22480 }, { "epoch": 3.24, "learning_rate": 1.0511659588108022e-05, "loss": 0.0004, "step": 22485 }, { "epoch": 3.24, "learning_rate": 1.0501642759836526e-05, "loss": 0.0006, "step": 22490 }, { "epoch": 3.24, "learning_rate": 1.049162593156503e-05, "loss": 0.0008, "step": 22495 }, { "epoch": 3.25, "learning_rate": 1.0481609103293534e-05, "loss": 0.001, "step": 22500 }, { "epoch": 3.25, "learning_rate": 1.0471592275022038e-05, "loss": 0.0008, "step": 22505 }, { "epoch": 3.25, "learning_rate": 1.0461575446750541e-05, "loss": 0.0008, "step": 22510 }, { "epoch": 3.25, "learning_rate": 1.0451558618479047e-05, "loss": 0.0005, "step": 22515 }, { "epoch": 3.25, "learning_rate": 1.0441541790207549e-05, "loss": 0.0005, "step": 22520 }, { "epoch": 3.25, "learning_rate": 1.0431524961936052e-05, "loss": 0.0008, "step": 22525 }, { "epoch": 3.25, "learning_rate": 1.0421508133664556e-05, "loss": 0.0006, "step": 22530 }, { "epoch": 3.25, "learning_rate": 1.0411491305393061e-05, "loss": 0.0008, "step": 22535 }, { "epoch": 3.25, "learning_rate": 1.0401474477121564e-05, "loss": 0.0008, "step": 22540 }, { "epoch": 3.25, "learning_rate": 1.0391457648850068e-05, "loss": 0.0006, "step": 22545 }, { "epoch": 3.25, "learning_rate": 1.0381440820578573e-05, "loss": 0.0004, "step": 22550 }, { "epoch": 3.25, "learning_rate": 1.0371423992307077e-05, "loss": 0.0008, "step": 22555 }, { "epoch": 3.25, "learning_rate": 1.036140716403558e-05, "loss": 0.0008, "step": 22560 }, { "epoch": 3.25, "learning_rate": 1.0351390335764084e-05, "loss": 0.0007, "step": 22565 }, { "epoch": 3.26, "learning_rate": 1.0341373507492589e-05, "loss": 0.0008, "step": 22570 }, { "epoch": 3.26, "learning_rate": 1.0331356679221093e-05, "loss": 0.001, "step": 22575 }, { "epoch": 3.26, "learning_rate": 1.0321339850949596e-05, "loss": 0.0007, "step": 22580 }, { "epoch": 3.26, "learning_rate": 1.03113230226781e-05, "loss": 0.0008, "step": 22585 }, { "epoch": 3.26, "learning_rate": 1.0301306194406605e-05, "loss": 0.0007, "step": 22590 }, { "epoch": 3.26, "learning_rate": 1.0291289366135108e-05, "loss": 0.0009, "step": 22595 }, { "epoch": 3.26, "learning_rate": 1.028127253786361e-05, "loss": 0.0007, "step": 22600 }, { "epoch": 3.26, "learning_rate": 1.0271255709592115e-05, "loss": 0.0005, "step": 22605 }, { "epoch": 3.26, "learning_rate": 1.0261238881320619e-05, "loss": 0.0006, "step": 22610 }, { "epoch": 3.26, "learning_rate": 1.0251222053049122e-05, "loss": 0.0007, "step": 22615 }, { "epoch": 3.26, "learning_rate": 1.0241205224777626e-05, "loss": 0.0007, "step": 22620 }, { "epoch": 3.26, "learning_rate": 1.0231188396506131e-05, "loss": 0.0009, "step": 22625 }, { "epoch": 3.26, "learning_rate": 1.0221171568234635e-05, "loss": 0.0004, "step": 22630 }, { "epoch": 3.26, "learning_rate": 1.0211154739963138e-05, "loss": 0.0004, "step": 22635 }, { "epoch": 3.27, "learning_rate": 1.0201137911691642e-05, "loss": 0.0008, "step": 22640 }, { "epoch": 3.27, "learning_rate": 1.0191121083420147e-05, "loss": 0.0007, "step": 22645 }, { "epoch": 3.27, "learning_rate": 1.018110425514865e-05, "loss": 0.0008, "step": 22650 }, { "epoch": 3.27, "learning_rate": 1.0171087426877154e-05, "loss": 0.0012, "step": 22655 }, { "epoch": 3.27, "learning_rate": 1.0161070598605658e-05, "loss": 0.0005, "step": 22660 }, { "epoch": 3.27, "learning_rate": 1.0151053770334163e-05, "loss": 0.0008, "step": 22665 }, { "epoch": 3.27, "learning_rate": 1.0141036942062666e-05, "loss": 0.0005, "step": 22670 }, { "epoch": 3.27, "learning_rate": 1.013102011379117e-05, "loss": 0.0007, "step": 22675 }, { "epoch": 3.27, "learning_rate": 1.0121003285519673e-05, "loss": 0.0005, "step": 22680 }, { "epoch": 3.27, "learning_rate": 1.0110986457248177e-05, "loss": 0.0005, "step": 22685 }, { "epoch": 3.27, "learning_rate": 1.010096962897668e-05, "loss": 0.0009, "step": 22690 }, { "epoch": 3.27, "learning_rate": 1.0090952800705184e-05, "loss": 0.0007, "step": 22695 }, { "epoch": 3.27, "learning_rate": 1.008093597243369e-05, "loss": 0.0008, "step": 22700 }, { "epoch": 3.27, "learning_rate": 1.0070919144162193e-05, "loss": 0.0008, "step": 22705 }, { "epoch": 3.28, "learning_rate": 1.0060902315890696e-05, "loss": 0.0009, "step": 22710 }, { "epoch": 3.28, "learning_rate": 1.00508854876192e-05, "loss": 0.0005, "step": 22715 }, { "epoch": 3.28, "learning_rate": 1.0040868659347705e-05, "loss": 0.0007, "step": 22720 }, { "epoch": 3.28, "learning_rate": 1.0030851831076209e-05, "loss": 0.0006, "step": 22725 }, { "epoch": 3.28, "learning_rate": 1.0020835002804712e-05, "loss": 0.0008, "step": 22730 }, { "epoch": 3.28, "learning_rate": 1.0010818174533217e-05, "loss": 0.0007, "step": 22735 }, { "epoch": 3.28, "learning_rate": 1.0000801346261721e-05, "loss": 0.001, "step": 22740 }, { "epoch": 3.28, "learning_rate": 9.990784517990224e-06, "loss": 0.0007, "step": 22745 }, { "epoch": 3.28, "learning_rate": 9.980767689718728e-06, "loss": 0.001, "step": 22750 }, { "epoch": 3.28, "learning_rate": 9.970750861447232e-06, "loss": 0.0008, "step": 22755 }, { "epoch": 3.28, "learning_rate": 9.960734033175735e-06, "loss": 0.0004, "step": 22760 }, { "epoch": 3.28, "learning_rate": 9.950717204904239e-06, "loss": 0.0008, "step": 22765 }, { "epoch": 3.28, "learning_rate": 9.940700376632744e-06, "loss": 0.0006, "step": 22770 }, { "epoch": 3.29, "learning_rate": 9.930683548361247e-06, "loss": 0.0008, "step": 22775 }, { "epoch": 3.29, "learning_rate": 9.920666720089751e-06, "loss": 0.0006, "step": 22780 }, { "epoch": 3.29, "learning_rate": 9.910649891818254e-06, "loss": 0.0004, "step": 22785 }, { "epoch": 3.29, "learning_rate": 9.90063306354676e-06, "loss": 0.0009, "step": 22790 }, { "epoch": 3.29, "learning_rate": 9.890616235275263e-06, "loss": 0.001, "step": 22795 }, { "epoch": 3.29, "learning_rate": 9.880599407003767e-06, "loss": 0.0008, "step": 22800 }, { "epoch": 3.29, "learning_rate": 9.87058257873227e-06, "loss": 0.0006, "step": 22805 }, { "epoch": 3.29, "learning_rate": 9.860565750460775e-06, "loss": 0.0008, "step": 22810 }, { "epoch": 3.29, "learning_rate": 9.850548922189279e-06, "loss": 0.0009, "step": 22815 }, { "epoch": 3.29, "learning_rate": 9.840532093917783e-06, "loss": 0.0007, "step": 22820 }, { "epoch": 3.29, "learning_rate": 9.830515265646286e-06, "loss": 0.0008, "step": 22825 }, { "epoch": 3.29, "learning_rate": 9.820498437374791e-06, "loss": 0.0008, "step": 22830 }, { "epoch": 3.29, "learning_rate": 9.810481609103293e-06, "loss": 0.0006, "step": 22835 }, { "epoch": 3.29, "learning_rate": 9.800464780831797e-06, "loss": 0.0006, "step": 22840 }, { "epoch": 3.3, "learning_rate": 9.790447952560302e-06, "loss": 0.0006, "step": 22845 }, { "epoch": 3.3, "learning_rate": 9.780431124288805e-06, "loss": 0.0009, "step": 22850 }, { "epoch": 3.3, "learning_rate": 9.770414296017309e-06, "loss": 0.0007, "step": 22855 }, { "epoch": 3.3, "learning_rate": 9.760397467745813e-06, "loss": 0.0007, "step": 22860 }, { "epoch": 3.3, "learning_rate": 9.750380639474318e-06, "loss": 0.0012, "step": 22865 }, { "epoch": 3.3, "learning_rate": 9.740363811202821e-06, "loss": 0.0007, "step": 22870 }, { "epoch": 3.3, "learning_rate": 9.730346982931325e-06, "loss": 0.0006, "step": 22875 }, { "epoch": 3.3, "learning_rate": 9.720330154659828e-06, "loss": 0.001, "step": 22880 }, { "epoch": 3.3, "learning_rate": 9.710313326388334e-06, "loss": 0.0009, "step": 22885 }, { "epoch": 3.3, "learning_rate": 9.700296498116837e-06, "loss": 0.0006, "step": 22890 }, { "epoch": 3.3, "learning_rate": 9.69027966984534e-06, "loss": 0.0008, "step": 22895 }, { "epoch": 3.3, "learning_rate": 9.680262841573844e-06, "loss": 0.0004, "step": 22900 }, { "epoch": 3.3, "learning_rate": 9.67024601330235e-06, "loss": 0.0005, "step": 22905 }, { "epoch": 3.3, "learning_rate": 9.660229185030853e-06, "loss": 0.0007, "step": 22910 }, { "epoch": 3.31, "learning_rate": 9.650212356759355e-06, "loss": 0.0005, "step": 22915 }, { "epoch": 3.31, "learning_rate": 9.64019552848786e-06, "loss": 0.0005, "step": 22920 }, { "epoch": 3.31, "learning_rate": 9.630178700216364e-06, "loss": 0.0007, "step": 22925 }, { "epoch": 3.31, "learning_rate": 9.620161871944867e-06, "loss": 0.0007, "step": 22930 }, { "epoch": 3.31, "learning_rate": 9.610145043673372e-06, "loss": 0.0007, "step": 22935 }, { "epoch": 3.31, "learning_rate": 9.600128215401876e-06, "loss": 0.0009, "step": 22940 }, { "epoch": 3.31, "learning_rate": 9.59011138713038e-06, "loss": 0.0007, "step": 22945 }, { "epoch": 3.31, "learning_rate": 9.580094558858883e-06, "loss": 0.0005, "step": 22950 }, { "epoch": 3.31, "learning_rate": 9.570077730587388e-06, "loss": 0.0007, "step": 22955 }, { "epoch": 3.31, "learning_rate": 9.560060902315892e-06, "loss": 0.0007, "step": 22960 }, { "epoch": 3.31, "learning_rate": 9.550044074044395e-06, "loss": 0.0005, "step": 22965 }, { "epoch": 3.31, "learning_rate": 9.540027245772899e-06, "loss": 0.0006, "step": 22970 }, { "epoch": 3.31, "learning_rate": 9.530010417501404e-06, "loss": 0.0005, "step": 22975 }, { "epoch": 3.31, "learning_rate": 9.519993589229907e-06, "loss": 0.0005, "step": 22980 }, { "epoch": 3.32, "learning_rate": 9.509976760958411e-06, "loss": 0.0006, "step": 22985 }, { "epoch": 3.32, "learning_rate": 9.499959932686915e-06, "loss": 0.0007, "step": 22990 }, { "epoch": 3.32, "learning_rate": 9.489943104415418e-06, "loss": 0.0007, "step": 22995 }, { "epoch": 3.32, "learning_rate": 9.479926276143922e-06, "loss": 0.0008, "step": 23000 }, { "epoch": 3.32, "learning_rate": 9.469909447872425e-06, "loss": 0.0008, "step": 23005 }, { "epoch": 3.32, "learning_rate": 9.45989261960093e-06, "loss": 0.0008, "step": 23010 }, { "epoch": 3.32, "learning_rate": 9.449875791329434e-06, "loss": 0.0005, "step": 23015 }, { "epoch": 3.32, "learning_rate": 9.439858963057937e-06, "loss": 0.0007, "step": 23020 }, { "epoch": 3.32, "learning_rate": 9.429842134786441e-06, "loss": 0.0008, "step": 23025 }, { "epoch": 3.32, "learning_rate": 9.419825306514946e-06, "loss": 0.0009, "step": 23030 }, { "epoch": 3.32, "learning_rate": 9.40980847824345e-06, "loss": 0.0007, "step": 23035 }, { "epoch": 3.32, "learning_rate": 9.399791649971953e-06, "loss": 0.0009, "step": 23040 }, { "epoch": 3.32, "learning_rate": 9.389774821700457e-06, "loss": 0.0008, "step": 23045 }, { "epoch": 3.32, "learning_rate": 9.379757993428962e-06, "loss": 0.001, "step": 23050 }, { "epoch": 3.33, "learning_rate": 9.369741165157466e-06, "loss": 0.0006, "step": 23055 }, { "epoch": 3.33, "learning_rate": 9.359724336885969e-06, "loss": 0.0005, "step": 23060 }, { "epoch": 3.33, "learning_rate": 9.349707508614473e-06, "loss": 0.0008, "step": 23065 }, { "epoch": 3.33, "learning_rate": 9.339690680342978e-06, "loss": 0.0009, "step": 23070 }, { "epoch": 3.33, "learning_rate": 9.32967385207148e-06, "loss": 0.0005, "step": 23075 }, { "epoch": 3.33, "learning_rate": 9.319657023799983e-06, "loss": 0.0009, "step": 23080 }, { "epoch": 3.33, "learning_rate": 9.309640195528488e-06, "loss": 0.0004, "step": 23085 }, { "epoch": 3.33, "learning_rate": 9.299623367256992e-06, "loss": 0.001, "step": 23090 }, { "epoch": 3.33, "learning_rate": 9.289606538985495e-06, "loss": 0.0009, "step": 23095 }, { "epoch": 3.33, "learning_rate": 9.279589710713999e-06, "loss": 0.0006, "step": 23100 }, { "epoch": 3.33, "learning_rate": 9.269572882442504e-06, "loss": 0.0005, "step": 23105 }, { "epoch": 3.33, "learning_rate": 9.259556054171008e-06, "loss": 0.0008, "step": 23110 }, { "epoch": 3.33, "learning_rate": 9.249539225899511e-06, "loss": 0.0005, "step": 23115 }, { "epoch": 3.33, "learning_rate": 9.239522397628017e-06, "loss": 0.0009, "step": 23120 }, { "epoch": 3.34, "learning_rate": 9.22950556935652e-06, "loss": 0.0007, "step": 23125 }, { "epoch": 3.34, "learning_rate": 9.219488741085024e-06, "loss": 0.0006, "step": 23130 }, { "epoch": 3.34, "learning_rate": 9.209471912813527e-06, "loss": 0.0007, "step": 23135 }, { "epoch": 3.34, "learning_rate": 9.199455084542032e-06, "loss": 0.0009, "step": 23140 }, { "epoch": 3.34, "learning_rate": 9.189438256270536e-06, "loss": 0.0009, "step": 23145 }, { "epoch": 3.34, "learning_rate": 9.179421427999038e-06, "loss": 0.0008, "step": 23150 }, { "epoch": 3.34, "learning_rate": 9.169404599727543e-06, "loss": 0.0006, "step": 23155 }, { "epoch": 3.34, "learning_rate": 9.159387771456046e-06, "loss": 0.0007, "step": 23160 }, { "epoch": 3.34, "learning_rate": 9.14937094318455e-06, "loss": 0.0006, "step": 23165 }, { "epoch": 3.34, "learning_rate": 9.139354114913054e-06, "loss": 0.0005, "step": 23170 }, { "epoch": 3.34, "learning_rate": 9.129337286641559e-06, "loss": 0.001, "step": 23175 }, { "epoch": 3.34, "learning_rate": 9.119320458370062e-06, "loss": 0.0011, "step": 23180 }, { "epoch": 3.34, "learning_rate": 9.109303630098566e-06, "loss": 0.0006, "step": 23185 }, { "epoch": 3.34, "learning_rate": 9.09928680182707e-06, "loss": 0.0009, "step": 23190 }, { "epoch": 3.35, "learning_rate": 9.089269973555575e-06, "loss": 0.0007, "step": 23195 }, { "epoch": 3.35, "learning_rate": 9.079253145284078e-06, "loss": 0.0006, "step": 23200 }, { "epoch": 3.35, "learning_rate": 9.069236317012582e-06, "loss": 0.0008, "step": 23205 }, { "epoch": 3.35, "learning_rate": 9.059219488741085e-06, "loss": 0.0007, "step": 23210 }, { "epoch": 3.35, "learning_rate": 9.04920266046959e-06, "loss": 0.0004, "step": 23215 }, { "epoch": 3.35, "learning_rate": 9.039185832198094e-06, "loss": 0.0009, "step": 23220 }, { "epoch": 3.35, "learning_rate": 9.029169003926597e-06, "loss": 0.0005, "step": 23225 }, { "epoch": 3.35, "learning_rate": 9.019152175655101e-06, "loss": 0.0006, "step": 23230 }, { "epoch": 3.35, "learning_rate": 9.009135347383605e-06, "loss": 0.0004, "step": 23235 }, { "epoch": 3.35, "learning_rate": 8.999118519112108e-06, "loss": 0.001, "step": 23240 }, { "epoch": 3.35, "learning_rate": 8.989101690840612e-06, "loss": 0.0008, "step": 23245 }, { "epoch": 3.35, "learning_rate": 8.979084862569117e-06, "loss": 0.0008, "step": 23250 }, { "epoch": 3.35, "learning_rate": 8.96906803429762e-06, "loss": 0.0005, "step": 23255 }, { "epoch": 3.35, "learning_rate": 8.959051206026124e-06, "loss": 0.0009, "step": 23260 }, { "epoch": 3.36, "learning_rate": 8.949034377754627e-06, "loss": 0.0007, "step": 23265 }, { "epoch": 3.36, "learning_rate": 8.939017549483133e-06, "loss": 0.0009, "step": 23270 }, { "epoch": 3.36, "learning_rate": 8.929000721211636e-06, "loss": 0.0009, "step": 23275 }, { "epoch": 3.36, "learning_rate": 8.91898389294014e-06, "loss": 0.0006, "step": 23280 }, { "epoch": 3.36, "learning_rate": 8.908967064668643e-06, "loss": 0.0009, "step": 23285 }, { "epoch": 3.36, "learning_rate": 8.898950236397149e-06, "loss": 0.0007, "step": 23290 }, { "epoch": 3.36, "learning_rate": 8.888933408125652e-06, "loss": 0.0005, "step": 23295 }, { "epoch": 3.36, "learning_rate": 8.878916579854156e-06, "loss": 0.0006, "step": 23300 }, { "epoch": 3.36, "learning_rate": 8.86889975158266e-06, "loss": 0.0006, "step": 23305 }, { "epoch": 3.36, "learning_rate": 8.858882923311163e-06, "loss": 0.0007, "step": 23310 }, { "epoch": 3.36, "learning_rate": 8.848866095039666e-06, "loss": 0.0008, "step": 23315 }, { "epoch": 3.36, "learning_rate": 8.83884926676817e-06, "loss": 0.0007, "step": 23320 }, { "epoch": 3.36, "learning_rate": 8.828832438496675e-06, "loss": 0.0005, "step": 23325 }, { "epoch": 3.37, "learning_rate": 8.818815610225178e-06, "loss": 0.001, "step": 23330 }, { "epoch": 3.37, "learning_rate": 8.808798781953682e-06, "loss": 0.0009, "step": 23335 }, { "epoch": 3.37, "learning_rate": 8.798781953682187e-06, "loss": 0.0007, "step": 23340 }, { "epoch": 3.37, "learning_rate": 8.78876512541069e-06, "loss": 0.001, "step": 23345 }, { "epoch": 3.37, "learning_rate": 8.778748297139194e-06, "loss": 0.0008, "step": 23350 }, { "epoch": 3.37, "learning_rate": 8.768731468867698e-06, "loss": 0.0006, "step": 23355 }, { "epoch": 3.37, "learning_rate": 8.758714640596203e-06, "loss": 0.0005, "step": 23360 }, { "epoch": 3.37, "learning_rate": 8.748697812324707e-06, "loss": 0.0009, "step": 23365 }, { "epoch": 3.37, "learning_rate": 8.73868098405321e-06, "loss": 0.0008, "step": 23370 }, { "epoch": 3.37, "learning_rate": 8.728664155781714e-06, "loss": 0.0007, "step": 23375 }, { "epoch": 3.37, "learning_rate": 8.718647327510219e-06, "loss": 0.0008, "step": 23380 }, { "epoch": 3.37, "learning_rate": 8.708630499238722e-06, "loss": 0.0007, "step": 23385 }, { "epoch": 3.37, "learning_rate": 8.698613670967224e-06, "loss": 0.0005, "step": 23390 }, { "epoch": 3.37, "learning_rate": 8.68859684269573e-06, "loss": 0.0008, "step": 23395 }, { "epoch": 3.38, "learning_rate": 8.678580014424233e-06, "loss": 0.0005, "step": 23400 }, { "epoch": 3.38, "learning_rate": 8.668563186152737e-06, "loss": 0.0008, "step": 23405 }, { "epoch": 3.38, "learning_rate": 8.65854635788124e-06, "loss": 0.0006, "step": 23410 }, { "epoch": 3.38, "learning_rate": 8.648529529609745e-06, "loss": 0.001, "step": 23415 }, { "epoch": 3.38, "learning_rate": 8.638512701338249e-06, "loss": 0.0006, "step": 23420 }, { "epoch": 3.38, "learning_rate": 8.628495873066752e-06, "loss": 0.0007, "step": 23425 }, { "epoch": 3.38, "learning_rate": 8.618479044795256e-06, "loss": 0.0005, "step": 23430 }, { "epoch": 3.38, "learning_rate": 8.608462216523761e-06, "loss": 0.0005, "step": 23435 }, { "epoch": 3.38, "learning_rate": 8.598445388252265e-06, "loss": 0.0006, "step": 23440 }, { "epoch": 3.38, "learning_rate": 8.588428559980768e-06, "loss": 0.0008, "step": 23445 }, { "epoch": 3.38, "learning_rate": 8.578411731709272e-06, "loss": 0.0007, "step": 23450 }, { "epoch": 3.38, "learning_rate": 8.568394903437777e-06, "loss": 0.001, "step": 23455 }, { "epoch": 3.38, "learning_rate": 8.55837807516628e-06, "loss": 0.0008, "step": 23460 }, { "epoch": 3.38, "learning_rate": 8.548361246894782e-06, "loss": 0.0007, "step": 23465 }, { "epoch": 3.39, "learning_rate": 8.538344418623288e-06, "loss": 0.0008, "step": 23470 }, { "epoch": 3.39, "learning_rate": 8.528327590351791e-06, "loss": 0.0006, "step": 23475 }, { "epoch": 3.39, "learning_rate": 8.518310762080295e-06, "loss": 0.0009, "step": 23480 }, { "epoch": 3.39, "learning_rate": 8.508293933808798e-06, "loss": 0.0004, "step": 23485 }, { "epoch": 3.39, "learning_rate": 8.498277105537303e-06, "loss": 0.0007, "step": 23490 }, { "epoch": 3.39, "learning_rate": 8.488260277265807e-06, "loss": 0.0005, "step": 23495 }, { "epoch": 3.39, "learning_rate": 8.47824344899431e-06, "loss": 0.0004, "step": 23500 }, { "epoch": 3.39, "learning_rate": 8.468226620722814e-06, "loss": 0.0005, "step": 23505 }, { "epoch": 3.39, "learning_rate": 8.45820979245132e-06, "loss": 0.0007, "step": 23510 }, { "epoch": 3.39, "learning_rate": 8.448192964179823e-06, "loss": 0.0007, "step": 23515 }, { "epoch": 3.39, "learning_rate": 8.438176135908326e-06, "loss": 0.001, "step": 23520 }, { "epoch": 3.39, "learning_rate": 8.428159307636831e-06, "loss": 0.0004, "step": 23525 }, { "epoch": 3.39, "learning_rate": 8.418142479365335e-06, "loss": 0.0007, "step": 23530 }, { "epoch": 3.39, "learning_rate": 8.408125651093839e-06, "loss": 0.0006, "step": 23535 }, { "epoch": 3.4, "learning_rate": 8.398108822822342e-06, "loss": 0.0005, "step": 23540 }, { "epoch": 3.4, "learning_rate": 8.388091994550846e-06, "loss": 0.0005, "step": 23545 }, { "epoch": 3.4, "learning_rate": 8.378075166279349e-06, "loss": 0.0007, "step": 23550 }, { "epoch": 3.4, "learning_rate": 8.368058338007853e-06, "loss": 0.0007, "step": 23555 }, { "epoch": 3.4, "learning_rate": 8.358041509736358e-06, "loss": 0.0007, "step": 23560 }, { "epoch": 3.4, "learning_rate": 8.348024681464861e-06, "loss": 0.0005, "step": 23565 }, { "epoch": 3.4, "learning_rate": 8.338007853193365e-06, "loss": 0.0008, "step": 23570 }, { "epoch": 3.4, "learning_rate": 8.327991024921868e-06, "loss": 0.0009, "step": 23575 }, { "epoch": 3.4, "learning_rate": 8.317974196650374e-06, "loss": 0.0006, "step": 23580 }, { "epoch": 3.4, "learning_rate": 8.307957368378877e-06, "loss": 0.001, "step": 23585 }, { "epoch": 3.4, "learning_rate": 8.29794054010738e-06, "loss": 0.0007, "step": 23590 }, { "epoch": 3.4, "learning_rate": 8.287923711835884e-06, "loss": 0.0011, "step": 23595 }, { "epoch": 3.4, "learning_rate": 8.27790688356439e-06, "loss": 0.0007, "step": 23600 }, { "epoch": 3.4, "learning_rate": 8.267890055292893e-06, "loss": 0.0006, "step": 23605 }, { "epoch": 3.41, "learning_rate": 8.257873227021397e-06, "loss": 0.0007, "step": 23610 }, { "epoch": 3.41, "learning_rate": 8.2478563987499e-06, "loss": 0.0005, "step": 23615 }, { "epoch": 3.41, "learning_rate": 8.237839570478405e-06, "loss": 0.001, "step": 23620 }, { "epoch": 3.41, "learning_rate": 8.227822742206907e-06, "loss": 0.0007, "step": 23625 }, { "epoch": 3.41, "learning_rate": 8.21780591393541e-06, "loss": 0.0006, "step": 23630 }, { "epoch": 3.41, "learning_rate": 8.207789085663916e-06, "loss": 0.0008, "step": 23635 }, { "epoch": 3.41, "learning_rate": 8.19777225739242e-06, "loss": 0.0007, "step": 23640 }, { "epoch": 3.41, "learning_rate": 8.187755429120923e-06, "loss": 0.0006, "step": 23645 }, { "epoch": 3.41, "learning_rate": 8.177738600849427e-06, "loss": 0.0007, "step": 23650 }, { "epoch": 3.41, "learning_rate": 8.167721772577932e-06, "loss": 0.0006, "step": 23655 }, { "epoch": 3.41, "learning_rate": 8.157704944306435e-06, "loss": 0.0009, "step": 23660 }, { "epoch": 3.41, "learning_rate": 8.147688116034939e-06, "loss": 0.0008, "step": 23665 }, { "epoch": 3.41, "learning_rate": 8.137671287763442e-06, "loss": 0.0007, "step": 23670 }, { "epoch": 3.41, "learning_rate": 8.127654459491948e-06, "loss": 0.0004, "step": 23675 }, { "epoch": 3.42, "learning_rate": 8.117637631220451e-06, "loss": 0.001, "step": 23680 }, { "epoch": 3.42, "learning_rate": 8.107620802948955e-06, "loss": 0.0006, "step": 23685 }, { "epoch": 3.42, "learning_rate": 8.097603974677458e-06, "loss": 0.0007, "step": 23690 }, { "epoch": 3.42, "learning_rate": 8.087587146405963e-06, "loss": 0.0007, "step": 23695 }, { "epoch": 3.42, "learning_rate": 8.077570318134467e-06, "loss": 0.0006, "step": 23700 }, { "epoch": 3.42, "learning_rate": 8.067553489862969e-06, "loss": 0.0008, "step": 23705 }, { "epoch": 3.42, "learning_rate": 8.057536661591474e-06, "loss": 0.0007, "step": 23710 }, { "epoch": 3.42, "learning_rate": 8.047519833319978e-06, "loss": 0.0009, "step": 23715 }, { "epoch": 3.42, "learning_rate": 8.037503005048481e-06, "loss": 0.0006, "step": 23720 }, { "epoch": 3.42, "learning_rate": 8.027486176776985e-06, "loss": 0.0005, "step": 23725 }, { "epoch": 3.42, "learning_rate": 8.01746934850549e-06, "loss": 0.0008, "step": 23730 }, { "epoch": 3.42, "learning_rate": 8.007452520233993e-06, "loss": 0.0006, "step": 23735 }, { "epoch": 3.42, "learning_rate": 7.997435691962497e-06, "loss": 0.0008, "step": 23740 }, { "epoch": 3.42, "learning_rate": 7.987418863691002e-06, "loss": 0.0005, "step": 23745 }, { "epoch": 3.43, "learning_rate": 7.977402035419506e-06, "loss": 0.0007, "step": 23750 }, { "epoch": 3.43, "learning_rate": 7.96738520714801e-06, "loss": 0.0008, "step": 23755 }, { "epoch": 3.43, "learning_rate": 7.957368378876513e-06, "loss": 0.0012, "step": 23760 }, { "epoch": 3.43, "learning_rate": 7.947351550605018e-06, "loss": 0.0006, "step": 23765 }, { "epoch": 3.43, "learning_rate": 7.937334722333522e-06, "loss": 0.0007, "step": 23770 }, { "epoch": 3.43, "learning_rate": 7.927317894062025e-06, "loss": 0.0012, "step": 23775 }, { "epoch": 3.43, "learning_rate": 7.917301065790529e-06, "loss": 0.0007, "step": 23780 }, { "epoch": 3.43, "learning_rate": 7.907284237519032e-06, "loss": 0.0006, "step": 23785 }, { "epoch": 3.43, "learning_rate": 7.897267409247536e-06, "loss": 0.0006, "step": 23790 }, { "epoch": 3.43, "learning_rate": 7.88725058097604e-06, "loss": 0.0006, "step": 23795 }, { "epoch": 3.43, "learning_rate": 7.877233752704544e-06, "loss": 0.0006, "step": 23800 }, { "epoch": 3.43, "learning_rate": 7.867216924433048e-06, "loss": 0.0009, "step": 23805 }, { "epoch": 3.43, "learning_rate": 7.857200096161551e-06, "loss": 0.0007, "step": 23810 }, { "epoch": 3.44, "learning_rate": 7.847183267890055e-06, "loss": 0.0008, "step": 23815 }, { "epoch": 3.44, "learning_rate": 7.83716643961856e-06, "loss": 0.0009, "step": 23820 }, { "epoch": 3.44, "learning_rate": 7.827149611347064e-06, "loss": 0.0004, "step": 23825 }, { "epoch": 3.44, "learning_rate": 7.817132783075567e-06, "loss": 0.001, "step": 23830 }, { "epoch": 3.44, "learning_rate": 7.80711595480407e-06, "loss": 0.0007, "step": 23835 }, { "epoch": 3.44, "learning_rate": 7.797099126532576e-06, "loss": 0.0007, "step": 23840 }, { "epoch": 3.44, "learning_rate": 7.78708229826108e-06, "loss": 0.0005, "step": 23845 }, { "epoch": 3.44, "learning_rate": 7.777065469989583e-06, "loss": 0.0011, "step": 23850 }, { "epoch": 3.44, "learning_rate": 7.767048641718087e-06, "loss": 0.0007, "step": 23855 }, { "epoch": 3.44, "learning_rate": 7.75703181344659e-06, "loss": 0.0006, "step": 23860 }, { "epoch": 3.44, "learning_rate": 7.747014985175094e-06, "loss": 0.0008, "step": 23865 }, { "epoch": 3.44, "learning_rate": 7.736998156903597e-06, "loss": 0.0006, "step": 23870 }, { "epoch": 3.44, "learning_rate": 7.726981328632102e-06, "loss": 0.0005, "step": 23875 }, { "epoch": 3.44, "learning_rate": 7.716964500360606e-06, "loss": 0.0007, "step": 23880 }, { "epoch": 3.45, "learning_rate": 7.70694767208911e-06, "loss": 0.0005, "step": 23885 }, { "epoch": 3.45, "learning_rate": 7.696930843817613e-06, "loss": 0.0008, "step": 23890 }, { "epoch": 3.45, "learning_rate": 7.686914015546118e-06, "loss": 0.001, "step": 23895 }, { "epoch": 3.45, "learning_rate": 7.676897187274622e-06, "loss": 0.0008, "step": 23900 }, { "epoch": 3.45, "learning_rate": 7.666880359003125e-06, "loss": 0.0007, "step": 23905 }, { "epoch": 3.45, "learning_rate": 7.656863530731629e-06, "loss": 0.001, "step": 23910 }, { "epoch": 3.45, "learning_rate": 7.646846702460134e-06, "loss": 0.0006, "step": 23915 }, { "epoch": 3.45, "learning_rate": 7.636829874188638e-06, "loss": 0.0007, "step": 23920 }, { "epoch": 3.45, "learning_rate": 7.626813045917142e-06, "loss": 0.0008, "step": 23925 }, { "epoch": 3.45, "learning_rate": 7.616796217645646e-06, "loss": 0.0007, "step": 23930 }, { "epoch": 3.45, "learning_rate": 7.60677938937415e-06, "loss": 0.0006, "step": 23935 }, { "epoch": 3.45, "learning_rate": 7.596762561102652e-06, "loss": 0.0008, "step": 23940 }, { "epoch": 3.45, "learning_rate": 7.586745732831156e-06, "loss": 0.0008, "step": 23945 }, { "epoch": 3.45, "learning_rate": 7.57672890455966e-06, "loss": 0.0008, "step": 23950 }, { "epoch": 3.46, "learning_rate": 7.566712076288164e-06, "loss": 0.0007, "step": 23955 }, { "epoch": 3.46, "learning_rate": 7.5566952480166685e-06, "loss": 0.0007, "step": 23960 }, { "epoch": 3.46, "learning_rate": 7.546678419745172e-06, "loss": 0.0006, "step": 23965 }, { "epoch": 3.46, "learning_rate": 7.536661591473676e-06, "loss": 0.0009, "step": 23970 }, { "epoch": 3.46, "learning_rate": 7.52664476320218e-06, "loss": 0.0007, "step": 23975 }, { "epoch": 3.46, "learning_rate": 7.516627934930684e-06, "loss": 0.0007, "step": 23980 }, { "epoch": 3.46, "learning_rate": 7.506611106659188e-06, "loss": 0.0009, "step": 23985 }, { "epoch": 3.46, "learning_rate": 7.496594278387692e-06, "loss": 0.0008, "step": 23990 }, { "epoch": 3.46, "learning_rate": 7.486577450116196e-06, "loss": 0.0006, "step": 23995 }, { "epoch": 3.46, "learning_rate": 7.4765606218447e-06, "loss": 0.0007, "step": 24000 }, { "epoch": 3.46, "learning_rate": 7.466543793573204e-06, "loss": 0.0012, "step": 24005 }, { "epoch": 3.46, "learning_rate": 7.456526965301708e-06, "loss": 0.0005, "step": 24010 }, { "epoch": 3.46, "learning_rate": 7.4465101370302116e-06, "loss": 0.0005, "step": 24015 }, { "epoch": 3.46, "learning_rate": 7.436493308758714e-06, "loss": 0.0006, "step": 24020 }, { "epoch": 3.47, "learning_rate": 7.426476480487219e-06, "loss": 0.0007, "step": 24025 }, { "epoch": 3.47, "learning_rate": 7.416459652215722e-06, "loss": 0.0007, "step": 24030 }, { "epoch": 3.47, "learning_rate": 7.4064428239442265e-06, "loss": 0.0009, "step": 24035 }, { "epoch": 3.47, "learning_rate": 7.39642599567273e-06, "loss": 0.0005, "step": 24040 }, { "epoch": 3.47, "learning_rate": 7.3864091674012345e-06, "loss": 0.0007, "step": 24045 }, { "epoch": 3.47, "learning_rate": 7.376392339129738e-06, "loss": 0.001, "step": 24050 }, { "epoch": 3.47, "learning_rate": 7.366375510858242e-06, "loss": 0.0005, "step": 24055 }, { "epoch": 3.47, "learning_rate": 7.356358682586746e-06, "loss": 0.0008, "step": 24060 }, { "epoch": 3.47, "learning_rate": 7.34634185431525e-06, "loss": 0.0006, "step": 24065 }, { "epoch": 3.47, "learning_rate": 7.336325026043754e-06, "loss": 0.0008, "step": 24070 }, { "epoch": 3.47, "learning_rate": 7.326308197772258e-06, "loss": 0.0004, "step": 24075 }, { "epoch": 3.47, "learning_rate": 7.316291369500762e-06, "loss": 0.0009, "step": 24080 }, { "epoch": 3.47, "learning_rate": 7.306274541229266e-06, "loss": 0.0008, "step": 24085 }, { "epoch": 3.47, "learning_rate": 7.29625771295777e-06, "loss": 0.0006, "step": 24090 }, { "epoch": 3.48, "learning_rate": 7.286240884686274e-06, "loss": 0.0009, "step": 24095 }, { "epoch": 3.48, "learning_rate": 7.276224056414777e-06, "loss": 0.0004, "step": 24100 }, { "epoch": 3.48, "learning_rate": 7.26620722814328e-06, "loss": 0.0008, "step": 24105 }, { "epoch": 3.48, "learning_rate": 7.256190399871785e-06, "loss": 0.0007, "step": 24110 }, { "epoch": 3.48, "learning_rate": 7.246173571600288e-06, "loss": 0.0007, "step": 24115 }, { "epoch": 3.48, "learning_rate": 7.2361567433287925e-06, "loss": 0.001, "step": 24120 }, { "epoch": 3.48, "learning_rate": 7.226139915057296e-06, "loss": 0.0009, "step": 24125 }, { "epoch": 3.48, "learning_rate": 7.2161230867858004e-06, "loss": 0.0005, "step": 24130 }, { "epoch": 3.48, "learning_rate": 7.206106258514304e-06, "loss": 0.0009, "step": 24135 }, { "epoch": 3.48, "learning_rate": 7.196089430242808e-06, "loss": 0.0008, "step": 24140 }, { "epoch": 3.48, "learning_rate": 7.186072601971313e-06, "loss": 0.0006, "step": 24145 }, { "epoch": 3.48, "learning_rate": 7.176055773699816e-06, "loss": 0.0008, "step": 24150 }, { "epoch": 3.48, "learning_rate": 7.166038945428321e-06, "loss": 0.0009, "step": 24155 }, { "epoch": 3.48, "learning_rate": 7.156022117156824e-06, "loss": 0.0008, "step": 24160 }, { "epoch": 3.49, "learning_rate": 7.1460052888853286e-06, "loss": 0.0005, "step": 24165 }, { "epoch": 3.49, "learning_rate": 7.135988460613832e-06, "loss": 0.0008, "step": 24170 }, { "epoch": 3.49, "learning_rate": 7.125971632342335e-06, "loss": 0.0004, "step": 24175 }, { "epoch": 3.49, "learning_rate": 7.115954804070839e-06, "loss": 0.0006, "step": 24180 }, { "epoch": 3.49, "learning_rate": 7.105937975799343e-06, "loss": 0.0006, "step": 24185 }, { "epoch": 3.49, "learning_rate": 7.095921147527847e-06, "loss": 0.0008, "step": 24190 }, { "epoch": 3.49, "learning_rate": 7.085904319256351e-06, "loss": 0.0005, "step": 24195 }, { "epoch": 3.49, "learning_rate": 7.075887490984855e-06, "loss": 0.0007, "step": 24200 }, { "epoch": 3.49, "learning_rate": 7.0658706627133585e-06, "loss": 0.0005, "step": 24205 }, { "epoch": 3.49, "learning_rate": 7.055853834441863e-06, "loss": 0.0009, "step": 24210 }, { "epoch": 3.49, "learning_rate": 7.045837006170366e-06, "loss": 0.0006, "step": 24215 }, { "epoch": 3.49, "learning_rate": 7.035820177898871e-06, "loss": 0.0007, "step": 24220 }, { "epoch": 3.49, "learning_rate": 7.025803349627374e-06, "loss": 0.001, "step": 24225 }, { "epoch": 3.49, "learning_rate": 7.015786521355879e-06, "loss": 0.0008, "step": 24230 }, { "epoch": 3.5, "learning_rate": 7.005769693084382e-06, "loss": 0.0011, "step": 24235 }, { "epoch": 3.5, "learning_rate": 6.995752864812887e-06, "loss": 0.0007, "step": 24240 }, { "epoch": 3.5, "learning_rate": 6.98573603654139e-06, "loss": 0.0009, "step": 24245 }, { "epoch": 3.5, "learning_rate": 6.9757192082698945e-06, "loss": 0.0006, "step": 24250 }, { "epoch": 3.5, "learning_rate": 6.965702379998397e-06, "loss": 0.0009, "step": 24255 }, { "epoch": 3.5, "learning_rate": 6.955685551726901e-06, "loss": 0.0008, "step": 24260 }, { "epoch": 3.5, "learning_rate": 6.945668723455405e-06, "loss": 0.001, "step": 24265 }, { "epoch": 3.5, "learning_rate": 6.935651895183909e-06, "loss": 0.0008, "step": 24270 }, { "epoch": 3.5, "learning_rate": 6.925635066912413e-06, "loss": 0.0012, "step": 24275 }, { "epoch": 3.5, "learning_rate": 6.915618238640917e-06, "loss": 0.0007, "step": 24280 }, { "epoch": 3.5, "learning_rate": 6.905601410369421e-06, "loss": 0.0005, "step": 24285 }, { "epoch": 3.5, "learning_rate": 6.8955845820979245e-06, "loss": 0.0008, "step": 24290 }, { "epoch": 3.5, "learning_rate": 6.885567753826429e-06, "loss": 0.0005, "step": 24295 }, { "epoch": 3.5, "learning_rate": 6.875550925554932e-06, "loss": 0.0007, "step": 24300 }, { "epoch": 3.51, "learning_rate": 6.865534097283437e-06, "loss": 0.0008, "step": 24305 }, { "epoch": 3.51, "learning_rate": 6.85551726901194e-06, "loss": 0.0008, "step": 24310 }, { "epoch": 3.51, "learning_rate": 6.845500440740445e-06, "loss": 0.0007, "step": 24315 }, { "epoch": 3.51, "learning_rate": 6.835483612468948e-06, "loss": 0.0005, "step": 24320 }, { "epoch": 3.51, "learning_rate": 6.825466784197453e-06, "loss": 0.0012, "step": 24325 }, { "epoch": 3.51, "learning_rate": 6.815449955925957e-06, "loss": 0.0009, "step": 24330 }, { "epoch": 3.51, "learning_rate": 6.805433127654459e-06, "loss": 0.0007, "step": 24335 }, { "epoch": 3.51, "learning_rate": 6.795416299382963e-06, "loss": 0.0007, "step": 24340 }, { "epoch": 3.51, "learning_rate": 6.785399471111467e-06, "loss": 0.0008, "step": 24345 }, { "epoch": 3.51, "learning_rate": 6.775382642839971e-06, "loss": 0.0005, "step": 24350 }, { "epoch": 3.51, "learning_rate": 6.765365814568475e-06, "loss": 0.0007, "step": 24355 }, { "epoch": 3.51, "learning_rate": 6.755348986296979e-06, "loss": 0.0007, "step": 24360 }, { "epoch": 3.51, "learning_rate": 6.745332158025483e-06, "loss": 0.0007, "step": 24365 }, { "epoch": 3.52, "learning_rate": 6.735315329753987e-06, "loss": 0.0004, "step": 24370 }, { "epoch": 3.52, "learning_rate": 6.725298501482491e-06, "loss": 0.0007, "step": 24375 }, { "epoch": 3.52, "learning_rate": 6.715281673210995e-06, "loss": 0.0007, "step": 24380 }, { "epoch": 3.52, "learning_rate": 6.705264844939499e-06, "loss": 0.0009, "step": 24385 }, { "epoch": 3.52, "learning_rate": 6.695248016668003e-06, "loss": 0.001, "step": 24390 }, { "epoch": 3.52, "learning_rate": 6.685231188396507e-06, "loss": 0.0009, "step": 24395 }, { "epoch": 3.52, "learning_rate": 6.675214360125011e-06, "loss": 0.0005, "step": 24400 }, { "epoch": 3.52, "learning_rate": 6.665197531853515e-06, "loss": 0.0005, "step": 24405 }, { "epoch": 3.52, "learning_rate": 6.655180703582019e-06, "loss": 0.0006, "step": 24410 }, { "epoch": 3.52, "learning_rate": 6.645163875310521e-06, "loss": 0.0008, "step": 24415 }, { "epoch": 3.52, "learning_rate": 6.635147047039026e-06, "loss": 0.0007, "step": 24420 }, { "epoch": 3.52, "learning_rate": 6.625130218767529e-06, "loss": 0.0008, "step": 24425 }, { "epoch": 3.52, "learning_rate": 6.6151133904960336e-06, "loss": 0.0006, "step": 24430 }, { "epoch": 3.52, "learning_rate": 6.605096562224537e-06, "loss": 0.0005, "step": 24435 }, { "epoch": 3.53, "learning_rate": 6.5950797339530415e-06, "loss": 0.0007, "step": 24440 }, { "epoch": 3.53, "learning_rate": 6.585062905681545e-06, "loss": 0.0005, "step": 24445 }, { "epoch": 3.53, "learning_rate": 6.575046077410049e-06, "loss": 0.001, "step": 24450 }, { "epoch": 3.53, "learning_rate": 6.565029249138553e-06, "loss": 0.001, "step": 24455 }, { "epoch": 3.53, "learning_rate": 6.555012420867057e-06, "loss": 0.0011, "step": 24460 }, { "epoch": 3.53, "learning_rate": 6.544995592595561e-06, "loss": 0.0005, "step": 24465 }, { "epoch": 3.53, "learning_rate": 6.534978764324065e-06, "loss": 0.0007, "step": 24470 }, { "epoch": 3.53, "learning_rate": 6.524961936052569e-06, "loss": 0.0007, "step": 24475 }, { "epoch": 3.53, "learning_rate": 6.514945107781073e-06, "loss": 0.0008, "step": 24480 }, { "epoch": 3.53, "learning_rate": 6.504928279509577e-06, "loss": 0.0008, "step": 24485 }, { "epoch": 3.53, "learning_rate": 6.494911451238079e-06, "loss": 0.0006, "step": 24490 }, { "epoch": 3.53, "learning_rate": 6.484894622966584e-06, "loss": 0.0006, "step": 24495 }, { "epoch": 3.53, "learning_rate": 6.474877794695087e-06, "loss": 0.0007, "step": 24500 }, { "epoch": 3.53, "learning_rate": 6.464860966423592e-06, "loss": 0.0011, "step": 24505 }, { "epoch": 3.54, "learning_rate": 6.454844138152095e-06, "loss": 0.001, "step": 24510 }, { "epoch": 3.54, "learning_rate": 6.4448273098805996e-06, "loss": 0.0009, "step": 24515 }, { "epoch": 3.54, "learning_rate": 6.434810481609103e-06, "loss": 0.0004, "step": 24520 }, { "epoch": 3.54, "learning_rate": 6.4247936533376075e-06, "loss": 0.001, "step": 24525 }, { "epoch": 3.54, "learning_rate": 6.414776825066111e-06, "loss": 0.0007, "step": 24530 }, { "epoch": 3.54, "learning_rate": 6.404759996794615e-06, "loss": 0.0005, "step": 24535 }, { "epoch": 3.54, "learning_rate": 6.394743168523119e-06, "loss": 0.0007, "step": 24540 }, { "epoch": 3.54, "learning_rate": 6.384726340251623e-06, "loss": 0.0006, "step": 24545 }, { "epoch": 3.54, "learning_rate": 6.374709511980128e-06, "loss": 0.0006, "step": 24550 }, { "epoch": 3.54, "learning_rate": 6.364692683708631e-06, "loss": 0.0008, "step": 24555 }, { "epoch": 3.54, "learning_rate": 6.354675855437136e-06, "loss": 0.0005, "step": 24560 }, { "epoch": 3.54, "learning_rate": 6.344659027165639e-06, "loss": 0.0004, "step": 24565 }, { "epoch": 3.54, "learning_rate": 6.334642198894142e-06, "loss": 0.0007, "step": 24570 }, { "epoch": 3.54, "learning_rate": 6.324625370622646e-06, "loss": 0.0009, "step": 24575 }, { "epoch": 3.55, "learning_rate": 6.31460854235115e-06, "loss": 0.0008, "step": 24580 }, { "epoch": 3.55, "learning_rate": 6.304591714079654e-06, "loss": 0.0008, "step": 24585 }, { "epoch": 3.55, "learning_rate": 6.294574885808158e-06, "loss": 0.0006, "step": 24590 }, { "epoch": 3.55, "learning_rate": 6.284558057536662e-06, "loss": 0.0006, "step": 24595 }, { "epoch": 3.55, "learning_rate": 6.2745412292651655e-06, "loss": 0.0005, "step": 24600 }, { "epoch": 3.55, "learning_rate": 6.26452440099367e-06, "loss": 0.0006, "step": 24605 }, { "epoch": 3.55, "learning_rate": 6.2545075727221735e-06, "loss": 0.0009, "step": 24610 }, { "epoch": 3.55, "learning_rate": 6.244490744450678e-06, "loss": 0.0007, "step": 24615 }, { "epoch": 3.55, "learning_rate": 6.234473916179181e-06, "loss": 0.0007, "step": 24620 }, { "epoch": 3.55, "learning_rate": 6.224457087907686e-06, "loss": 0.0008, "step": 24625 }, { "epoch": 3.55, "learning_rate": 6.2144402596361884e-06, "loss": 0.0008, "step": 24630 }, { "epoch": 3.55, "learning_rate": 6.204423431364693e-06, "loss": 0.0007, "step": 24635 }, { "epoch": 3.55, "learning_rate": 6.194406603093196e-06, "loss": 0.0007, "step": 24640 }, { "epoch": 3.55, "learning_rate": 6.184389774821701e-06, "loss": 0.0006, "step": 24645 }, { "epoch": 3.56, "learning_rate": 6.174372946550204e-06, "loss": 0.0008, "step": 24650 }, { "epoch": 3.56, "learning_rate": 6.164356118278709e-06, "loss": 0.0006, "step": 24655 }, { "epoch": 3.56, "learning_rate": 6.154339290007213e-06, "loss": 0.0005, "step": 24660 }, { "epoch": 3.56, "learning_rate": 6.1443224617357166e-06, "loss": 0.0005, "step": 24665 }, { "epoch": 3.56, "learning_rate": 6.13430563346422e-06, "loss": 0.0011, "step": 24670 }, { "epoch": 3.56, "learning_rate": 6.124288805192724e-06, "loss": 0.0008, "step": 24675 }, { "epoch": 3.56, "learning_rate": 6.114271976921228e-06, "loss": 0.0006, "step": 24680 }, { "epoch": 3.56, "learning_rate": 6.1042551486497315e-06, "loss": 0.0008, "step": 24685 }, { "epoch": 3.56, "learning_rate": 6.094238320378236e-06, "loss": 0.0006, "step": 24690 }, { "epoch": 3.56, "learning_rate": 6.0842214921067394e-06, "loss": 0.0006, "step": 24695 }, { "epoch": 3.56, "learning_rate": 6.074204663835244e-06, "loss": 0.0005, "step": 24700 }, { "epoch": 3.56, "learning_rate": 6.064187835563747e-06, "loss": 0.0006, "step": 24705 }, { "epoch": 3.56, "learning_rate": 6.054171007292251e-06, "loss": 0.0009, "step": 24710 }, { "epoch": 3.56, "learning_rate": 6.044154179020755e-06, "loss": 0.0004, "step": 24715 }, { "epoch": 3.57, "learning_rate": 6.034137350749259e-06, "loss": 0.0009, "step": 24720 }, { "epoch": 3.57, "learning_rate": 6.024120522477763e-06, "loss": 0.0009, "step": 24725 }, { "epoch": 3.57, "learning_rate": 6.014103694206267e-06, "loss": 0.0008, "step": 24730 }, { "epoch": 3.57, "learning_rate": 6.004086865934771e-06, "loss": 0.001, "step": 24735 }, { "epoch": 3.57, "learning_rate": 5.994070037663275e-06, "loss": 0.0007, "step": 24740 }, { "epoch": 3.57, "learning_rate": 5.984053209391779e-06, "loss": 0.0005, "step": 24745 }, { "epoch": 3.57, "learning_rate": 5.974036381120282e-06, "loss": 0.0007, "step": 24750 }, { "epoch": 3.57, "learning_rate": 5.964019552848786e-06, "loss": 0.0008, "step": 24755 }, { "epoch": 3.57, "learning_rate": 5.9540027245772904e-06, "loss": 0.0009, "step": 24760 }, { "epoch": 3.57, "learning_rate": 5.943985896305794e-06, "loss": 0.0012, "step": 24765 }, { "epoch": 3.57, "learning_rate": 5.933969068034298e-06, "loss": 0.0007, "step": 24770 }, { "epoch": 3.57, "learning_rate": 5.923952239762802e-06, "loss": 0.0008, "step": 24775 }, { "epoch": 3.57, "learning_rate": 5.913935411491306e-06, "loss": 0.0008, "step": 24780 }, { "epoch": 3.57, "learning_rate": 5.90391858321981e-06, "loss": 0.0007, "step": 24785 }, { "epoch": 3.58, "learning_rate": 5.893901754948313e-06, "loss": 0.0005, "step": 24790 }, { "epoch": 3.58, "learning_rate": 5.883884926676817e-06, "loss": 0.0009, "step": 24795 }, { "epoch": 3.58, "learning_rate": 5.873868098405321e-06, "loss": 0.0006, "step": 24800 }, { "epoch": 3.58, "learning_rate": 5.863851270133825e-06, "loss": 0.0007, "step": 24805 }, { "epoch": 3.58, "learning_rate": 5.853834441862329e-06, "loss": 0.0009, "step": 24810 }, { "epoch": 3.58, "learning_rate": 5.843817613590833e-06, "loss": 0.0006, "step": 24815 }, { "epoch": 3.58, "learning_rate": 5.833800785319337e-06, "loss": 0.0007, "step": 24820 }, { "epoch": 3.58, "learning_rate": 5.823783957047841e-06, "loss": 0.0008, "step": 24825 }, { "epoch": 3.58, "learning_rate": 5.813767128776344e-06, "loss": 0.0009, "step": 24830 }, { "epoch": 3.58, "learning_rate": 5.8037503005048485e-06, "loss": 0.0004, "step": 24835 }, { "epoch": 3.58, "learning_rate": 5.793733472233352e-06, "loss": 0.0007, "step": 24840 }, { "epoch": 3.58, "learning_rate": 5.7837166439618564e-06, "loss": 0.0009, "step": 24845 }, { "epoch": 3.58, "learning_rate": 5.77369981569036e-06, "loss": 0.001, "step": 24850 }, { "epoch": 3.59, "learning_rate": 5.763682987418864e-06, "loss": 0.0012, "step": 24855 }, { "epoch": 3.59, "learning_rate": 5.753666159147368e-06, "loss": 0.0008, "step": 24860 }, { "epoch": 3.59, "learning_rate": 5.743649330875871e-06, "loss": 0.001, "step": 24865 }, { "epoch": 3.59, "learning_rate": 5.733632502604376e-06, "loss": 0.0007, "step": 24870 }, { "epoch": 3.59, "learning_rate": 5.723615674332879e-06, "loss": 0.001, "step": 24875 }, { "epoch": 3.59, "learning_rate": 5.713598846061384e-06, "loss": 0.0011, "step": 24880 }, { "epoch": 3.59, "learning_rate": 5.703582017789887e-06, "loss": 0.0005, "step": 24885 }, { "epoch": 3.59, "learning_rate": 5.693565189518392e-06, "loss": 0.0007, "step": 24890 }, { "epoch": 3.59, "learning_rate": 5.683548361246895e-06, "loss": 0.0006, "step": 24895 }, { "epoch": 3.59, "learning_rate": 5.6735315329753995e-06, "loss": 0.0006, "step": 24900 }, { "epoch": 3.59, "learning_rate": 5.663514704703902e-06, "loss": 0.0011, "step": 24905 }, { "epoch": 3.59, "learning_rate": 5.653497876432407e-06, "loss": 0.0007, "step": 24910 }, { "epoch": 3.59, "learning_rate": 5.64348104816091e-06, "loss": 0.0008, "step": 24915 }, { "epoch": 3.59, "learning_rate": 5.6334642198894145e-06, "loss": 0.0008, "step": 24920 }, { "epoch": 3.6, "learning_rate": 5.623447391617918e-06, "loss": 0.0007, "step": 24925 }, { "epoch": 3.6, "learning_rate": 5.613430563346422e-06, "loss": 0.0007, "step": 24930 }, { "epoch": 3.6, "learning_rate": 5.603413735074926e-06, "loss": 0.0011, "step": 24935 }, { "epoch": 3.6, "learning_rate": 5.59339690680343e-06, "loss": 0.0006, "step": 24940 }, { "epoch": 3.6, "learning_rate": 5.583380078531934e-06, "loss": 0.0006, "step": 24945 }, { "epoch": 3.6, "learning_rate": 5.573363250260437e-06, "loss": 0.0003, "step": 24950 }, { "epoch": 3.6, "learning_rate": 5.563346421988942e-06, "loss": 0.0006, "step": 24955 }, { "epoch": 3.6, "learning_rate": 5.553329593717445e-06, "loss": 0.0007, "step": 24960 }, { "epoch": 3.6, "learning_rate": 5.54331276544595e-06, "loss": 0.0007, "step": 24965 }, { "epoch": 3.6, "learning_rate": 5.533295937174453e-06, "loss": 0.0011, "step": 24970 }, { "epoch": 3.6, "learning_rate": 5.523279108902958e-06, "loss": 0.0006, "step": 24975 }, { "epoch": 3.6, "learning_rate": 5.513262280631461e-06, "loss": 0.0009, "step": 24980 }, { "epoch": 3.6, "learning_rate": 5.503245452359965e-06, "loss": 0.0007, "step": 24985 }, { "epoch": 3.6, "learning_rate": 5.493228624088469e-06, "loss": 0.0006, "step": 24990 }, { "epoch": 3.61, "learning_rate": 5.483211795816973e-06, "loss": 0.0007, "step": 24995 }, { "epoch": 3.61, "learning_rate": 5.473194967545477e-06, "loss": 0.0007, "step": 25000 }, { "epoch": 3.61, "learning_rate": 5.4631781392739805e-06, "loss": 0.0008, "step": 25005 }, { "epoch": 3.61, "learning_rate": 5.453161311002485e-06, "loss": 0.0008, "step": 25010 }, { "epoch": 3.61, "learning_rate": 5.443144482730988e-06, "loss": 0.0008, "step": 25015 }, { "epoch": 3.61, "learning_rate": 5.433127654459493e-06, "loss": 0.0007, "step": 25020 }, { "epoch": 3.61, "learning_rate": 5.4231108261879955e-06, "loss": 0.0005, "step": 25025 }, { "epoch": 3.61, "learning_rate": 5.4130939979165e-06, "loss": 0.0009, "step": 25030 }, { "epoch": 3.61, "learning_rate": 5.403077169645003e-06, "loss": 0.0006, "step": 25035 }, { "epoch": 3.61, "learning_rate": 5.393060341373508e-06, "loss": 0.0007, "step": 25040 }, { "epoch": 3.61, "learning_rate": 5.383043513102011e-06, "loss": 0.0008, "step": 25045 }, { "epoch": 3.61, "learning_rate": 5.373026684830516e-06, "loss": 0.0009, "step": 25050 }, { "epoch": 3.61, "learning_rate": 5.36300985655902e-06, "loss": 0.0007, "step": 25055 }, { "epoch": 3.61, "learning_rate": 5.352993028287524e-06, "loss": 0.0006, "step": 25060 }, { "epoch": 3.62, "learning_rate": 5.342976200016027e-06, "loss": 0.0009, "step": 25065 }, { "epoch": 3.62, "learning_rate": 5.332959371744531e-06, "loss": 0.0006, "step": 25070 }, { "epoch": 3.62, "learning_rate": 5.322942543473035e-06, "loss": 0.0005, "step": 25075 }, { "epoch": 3.62, "learning_rate": 5.3129257152015386e-06, "loss": 0.0005, "step": 25080 }, { "epoch": 3.62, "learning_rate": 5.302908886930043e-06, "loss": 0.001, "step": 25085 }, { "epoch": 3.62, "learning_rate": 5.2928920586585465e-06, "loss": 0.0006, "step": 25090 }, { "epoch": 3.62, "learning_rate": 5.282875230387051e-06, "loss": 0.0005, "step": 25095 }, { "epoch": 3.62, "learning_rate": 5.272858402115554e-06, "loss": 0.0005, "step": 25100 }, { "epoch": 3.62, "learning_rate": 5.262841573844058e-06, "loss": 0.0004, "step": 25105 }, { "epoch": 3.62, "learning_rate": 5.252824745572562e-06, "loss": 0.0005, "step": 25110 }, { "epoch": 3.62, "learning_rate": 5.242807917301066e-06, "loss": 0.0007, "step": 25115 }, { "epoch": 3.62, "learning_rate": 5.23279108902957e-06, "loss": 0.0006, "step": 25120 }, { "epoch": 3.62, "learning_rate": 5.222774260758074e-06, "loss": 0.0007, "step": 25125 }, { "epoch": 3.62, "learning_rate": 5.212757432486578e-06, "loss": 0.0008, "step": 25130 }, { "epoch": 3.63, "learning_rate": 5.202740604215082e-06, "loss": 0.0005, "step": 25135 }, { "epoch": 3.63, "learning_rate": 5.192723775943586e-06, "loss": 0.0008, "step": 25140 }, { "epoch": 3.63, "learning_rate": 5.182706947672089e-06, "loss": 0.0008, "step": 25145 }, { "epoch": 3.63, "learning_rate": 5.172690119400593e-06, "loss": 0.001, "step": 25150 }, { "epoch": 3.63, "learning_rate": 5.162673291129097e-06, "loss": 0.0011, "step": 25155 }, { "epoch": 3.63, "learning_rate": 5.152656462857601e-06, "loss": 0.0012, "step": 25160 }, { "epoch": 3.63, "learning_rate": 5.142639634586105e-06, "loss": 0.0004, "step": 25165 }, { "epoch": 3.63, "learning_rate": 5.132622806314609e-06, "loss": 0.0008, "step": 25170 }, { "epoch": 3.63, "learning_rate": 5.122605978043113e-06, "loss": 0.0005, "step": 25175 }, { "epoch": 3.63, "learning_rate": 5.112589149771617e-06, "loss": 0.0006, "step": 25180 }, { "epoch": 3.63, "learning_rate": 5.10257232150012e-06, "loss": 0.0006, "step": 25185 }, { "epoch": 3.63, "learning_rate": 5.092555493228624e-06, "loss": 0.0008, "step": 25190 }, { "epoch": 3.63, "learning_rate": 5.082538664957128e-06, "loss": 0.0006, "step": 25195 }, { "epoch": 3.63, "learning_rate": 5.072521836685632e-06, "loss": 0.0009, "step": 25200 }, { "epoch": 3.64, "learning_rate": 5.062505008414136e-06, "loss": 0.0008, "step": 25205 }, { "epoch": 3.64, "learning_rate": 5.05248818014264e-06, "loss": 0.0004, "step": 25210 }, { "epoch": 3.64, "learning_rate": 5.042471351871144e-06, "loss": 0.0005, "step": 25215 }, { "epoch": 3.64, "learning_rate": 5.032454523599648e-06, "loss": 0.0009, "step": 25220 }, { "epoch": 3.64, "learning_rate": 5.022437695328151e-06, "loss": 0.0005, "step": 25225 }, { "epoch": 3.64, "learning_rate": 5.0124208670566556e-06, "loss": 0.0011, "step": 25230 }, { "epoch": 3.64, "learning_rate": 5.002404038785159e-06, "loss": 0.0005, "step": 25235 }, { "epoch": 3.64, "learning_rate": 4.9923872105136635e-06, "loss": 0.0009, "step": 25240 }, { "epoch": 3.64, "learning_rate": 4.982370382242167e-06, "loss": 0.0006, "step": 25245 }, { "epoch": 3.64, "learning_rate": 4.972353553970671e-06, "loss": 0.0007, "step": 25250 }, { "epoch": 3.64, "learning_rate": 4.962336725699175e-06, "loss": 0.0005, "step": 25255 }, { "epoch": 3.64, "learning_rate": 4.9523198974276784e-06, "loss": 0.001, "step": 25260 }, { "epoch": 3.64, "learning_rate": 4.942303069156182e-06, "loss": 0.0013, "step": 25265 }, { "epoch": 3.64, "learning_rate": 4.932286240884686e-06, "loss": 0.0008, "step": 25270 }, { "epoch": 3.65, "learning_rate": 4.922269412613191e-06, "loss": 0.0009, "step": 25275 }, { "epoch": 3.65, "learning_rate": 4.912252584341694e-06, "loss": 0.0009, "step": 25280 }, { "epoch": 3.65, "learning_rate": 4.902235756070199e-06, "loss": 0.0011, "step": 25285 }, { "epoch": 3.65, "learning_rate": 4.892218927798702e-06, "loss": 0.0008, "step": 25290 }, { "epoch": 3.65, "learning_rate": 4.8822020995272066e-06, "loss": 0.0007, "step": 25295 }, { "epoch": 3.65, "learning_rate": 4.872185271255709e-06, "loss": 0.0006, "step": 25300 }, { "epoch": 3.65, "learning_rate": 4.862168442984214e-06, "loss": 0.0007, "step": 25305 }, { "epoch": 3.65, "learning_rate": 4.852151614712717e-06, "loss": 0.0008, "step": 25310 }, { "epoch": 3.65, "learning_rate": 4.8421347864412215e-06, "loss": 0.0004, "step": 25315 }, { "epoch": 3.65, "learning_rate": 4.832117958169725e-06, "loss": 0.0006, "step": 25320 }, { "epoch": 3.65, "learning_rate": 4.8221011298982295e-06, "loss": 0.0009, "step": 25325 }, { "epoch": 3.65, "learning_rate": 4.812084301626733e-06, "loss": 0.0007, "step": 25330 }, { "epoch": 3.65, "learning_rate": 4.802067473355237e-06, "loss": 0.0007, "step": 25335 }, { "epoch": 3.65, "learning_rate": 4.792050645083741e-06, "loss": 0.0008, "step": 25340 }, { "epoch": 3.66, "learning_rate": 4.7820338168122444e-06, "loss": 0.0008, "step": 25345 }, { "epoch": 3.66, "learning_rate": 4.772016988540749e-06, "loss": 0.0005, "step": 25350 }, { "epoch": 3.66, "learning_rate": 4.762000160269252e-06, "loss": 0.0005, "step": 25355 }, { "epoch": 3.66, "learning_rate": 4.751983331997757e-06, "loss": 0.0008, "step": 25360 }, { "epoch": 3.66, "learning_rate": 4.74196650372626e-06, "loss": 0.0005, "step": 25365 }, { "epoch": 3.66, "learning_rate": 4.731949675454765e-06, "loss": 0.0008, "step": 25370 }, { "epoch": 3.66, "learning_rate": 4.721932847183268e-06, "loss": 0.0009, "step": 25375 }, { "epoch": 3.66, "learning_rate": 4.711916018911772e-06, "loss": 0.0009, "step": 25380 }, { "epoch": 3.66, "learning_rate": 4.701899190640276e-06, "loss": 0.0008, "step": 25385 }, { "epoch": 3.66, "learning_rate": 4.69188236236878e-06, "loss": 0.0009, "step": 25390 }, { "epoch": 3.66, "learning_rate": 4.681865534097284e-06, "loss": 0.0007, "step": 25395 }, { "epoch": 3.66, "learning_rate": 4.6718487058257875e-06, "loss": 0.0006, "step": 25400 }, { "epoch": 3.66, "learning_rate": 4.661831877554292e-06, "loss": 0.0006, "step": 25405 }, { "epoch": 3.67, "learning_rate": 4.6518150492827954e-06, "loss": 0.0006, "step": 25410 }, { "epoch": 3.67, "learning_rate": 4.6417982210113e-06, "loss": 0.0007, "step": 25415 }, { "epoch": 3.67, "learning_rate": 4.6317813927398025e-06, "loss": 0.0007, "step": 25420 }, { "epoch": 3.67, "learning_rate": 4.621764564468307e-06, "loss": 0.0006, "step": 25425 }, { "epoch": 3.67, "learning_rate": 4.61174773619681e-06, "loss": 0.0007, "step": 25430 }, { "epoch": 3.67, "learning_rate": 4.601730907925315e-06, "loss": 0.0007, "step": 25435 }, { "epoch": 3.67, "learning_rate": 4.591714079653818e-06, "loss": 0.0011, "step": 25440 }, { "epoch": 3.67, "learning_rate": 4.581697251382323e-06, "loss": 0.0009, "step": 25445 }, { "epoch": 3.67, "learning_rate": 4.571680423110826e-06, "loss": 0.0006, "step": 25450 }, { "epoch": 3.67, "learning_rate": 4.561663594839331e-06, "loss": 0.0006, "step": 25455 }, { "epoch": 3.67, "learning_rate": 4.551646766567834e-06, "loss": 0.0009, "step": 25460 }, { "epoch": 3.67, "learning_rate": 4.541629938296338e-06, "loss": 0.001, "step": 25465 }, { "epoch": 3.67, "learning_rate": 4.531613110024842e-06, "loss": 0.0006, "step": 25470 }, { "epoch": 3.67, "learning_rate": 4.521596281753346e-06, "loss": 0.0005, "step": 25475 }, { "epoch": 3.68, "learning_rate": 4.51157945348185e-06, "loss": 0.0006, "step": 25480 }, { "epoch": 3.68, "learning_rate": 4.5015626252103535e-06, "loss": 0.0006, "step": 25485 }, { "epoch": 3.68, "learning_rate": 4.491545796938858e-06, "loss": 0.0008, "step": 25490 }, { "epoch": 3.68, "learning_rate": 4.481528968667361e-06, "loss": 0.0006, "step": 25495 }, { "epoch": 3.68, "learning_rate": 4.471512140395865e-06, "loss": 0.0007, "step": 25500 }, { "epoch": 3.68, "learning_rate": 4.461495312124369e-06, "loss": 0.0008, "step": 25505 }, { "epoch": 3.68, "learning_rate": 4.451478483852873e-06, "loss": 0.0007, "step": 25510 }, { "epoch": 3.68, "learning_rate": 4.441461655581377e-06, "loss": 0.0007, "step": 25515 }, { "epoch": 3.68, "learning_rate": 4.431444827309881e-06, "loss": 0.0006, "step": 25520 }, { "epoch": 3.68, "learning_rate": 4.421427999038385e-06, "loss": 0.0007, "step": 25525 }, { "epoch": 3.68, "learning_rate": 4.411411170766889e-06, "loss": 0.001, "step": 25530 }, { "epoch": 3.68, "learning_rate": 4.401394342495393e-06, "loss": 0.0007, "step": 25535 }, { "epoch": 3.68, "learning_rate": 4.391377514223896e-06, "loss": 0.0007, "step": 25540 }, { "epoch": 3.68, "learning_rate": 4.3813606859524e-06, "loss": 0.0008, "step": 25545 }, { "epoch": 3.69, "learning_rate": 4.371343857680904e-06, "loss": 0.0007, "step": 25550 }, { "epoch": 3.69, "learning_rate": 4.361327029409408e-06, "loss": 0.0007, "step": 25555 }, { "epoch": 3.69, "learning_rate": 4.351310201137912e-06, "loss": 0.0011, "step": 25560 }, { "epoch": 3.69, "learning_rate": 4.341293372866416e-06, "loss": 0.0015, "step": 25565 }, { "epoch": 3.69, "learning_rate": 4.33127654459492e-06, "loss": 0.0006, "step": 25570 }, { "epoch": 3.69, "learning_rate": 4.321259716323423e-06, "loss": 0.0004, "step": 25575 }, { "epoch": 3.69, "learning_rate": 4.311242888051927e-06, "loss": 0.0006, "step": 25580 }, { "epoch": 3.69, "learning_rate": 4.301226059780431e-06, "loss": 0.0006, "step": 25585 }, { "epoch": 3.69, "learning_rate": 4.291209231508935e-06, "loss": 0.0006, "step": 25590 }, { "epoch": 3.69, "learning_rate": 4.281192403237439e-06, "loss": 0.001, "step": 25595 }, { "epoch": 3.69, "learning_rate": 4.271175574965943e-06, "loss": 0.0008, "step": 25600 }, { "epoch": 3.69, "learning_rate": 4.261158746694447e-06, "loss": 0.0008, "step": 25605 }, { "epoch": 3.69, "learning_rate": 4.251141918422951e-06, "loss": 0.0009, "step": 25610 }, { "epoch": 3.69, "learning_rate": 4.241125090151455e-06, "loss": 0.001, "step": 25615 }, { "epoch": 3.7, "learning_rate": 4.231108261879958e-06, "loss": 0.0005, "step": 25620 }, { "epoch": 3.7, "learning_rate": 4.221091433608463e-06, "loss": 0.0008, "step": 25625 }, { "epoch": 3.7, "learning_rate": 4.211074605336966e-06, "loss": 0.0007, "step": 25630 }, { "epoch": 3.7, "learning_rate": 4.2010577770654705e-06, "loss": 0.0009, "step": 25635 }, { "epoch": 3.7, "learning_rate": 4.191040948793974e-06, "loss": 0.0011, "step": 25640 }, { "epoch": 3.7, "learning_rate": 4.181024120522478e-06, "loss": 0.0007, "step": 25645 }, { "epoch": 3.7, "learning_rate": 4.171007292250982e-06, "loss": 0.0008, "step": 25650 }, { "epoch": 3.7, "learning_rate": 4.1609904639794855e-06, "loss": 0.0011, "step": 25655 }, { "epoch": 3.7, "learning_rate": 4.150973635707989e-06, "loss": 0.0006, "step": 25660 }, { "epoch": 3.7, "learning_rate": 4.140956807436493e-06, "loss": 0.0006, "step": 25665 }, { "epoch": 3.7, "learning_rate": 4.130939979164997e-06, "loss": 0.0004, "step": 25670 }, { "epoch": 3.7, "learning_rate": 4.120923150893501e-06, "loss": 0.0008, "step": 25675 }, { "epoch": 3.7, "learning_rate": 4.110906322622006e-06, "loss": 0.0008, "step": 25680 }, { "epoch": 3.7, "learning_rate": 4.100889494350509e-06, "loss": 0.0007, "step": 25685 }, { "epoch": 3.71, "learning_rate": 4.090872666079014e-06, "loss": 0.0008, "step": 25690 }, { "epoch": 3.71, "learning_rate": 4.080855837807516e-06, "loss": 0.0006, "step": 25695 }, { "epoch": 3.71, "learning_rate": 4.070839009536021e-06, "loss": 0.0008, "step": 25700 }, { "epoch": 3.71, "learning_rate": 4.060822181264524e-06, "loss": 0.0007, "step": 25705 }, { "epoch": 3.71, "learning_rate": 4.0508053529930286e-06, "loss": 0.001, "step": 25710 }, { "epoch": 3.71, "learning_rate": 4.040788524721532e-06, "loss": 0.0009, "step": 25715 }, { "epoch": 3.71, "learning_rate": 4.0307716964500365e-06, "loss": 0.0005, "step": 25720 }, { "epoch": 3.71, "learning_rate": 4.02075486817854e-06, "loss": 0.0008, "step": 25725 }, { "epoch": 3.71, "learning_rate": 4.010738039907044e-06, "loss": 0.001, "step": 25730 }, { "epoch": 3.71, "learning_rate": 4.000721211635548e-06, "loss": 0.0006, "step": 25735 }, { "epoch": 3.71, "learning_rate": 3.9907043833640515e-06, "loss": 0.0006, "step": 25740 }, { "epoch": 3.71, "learning_rate": 3.980687555092556e-06, "loss": 0.0006, "step": 25745 }, { "epoch": 3.71, "learning_rate": 3.970670726821059e-06, "loss": 0.0008, "step": 25750 }, { "epoch": 3.71, "learning_rate": 3.960653898549564e-06, "loss": 0.001, "step": 25755 }, { "epoch": 3.72, "learning_rate": 3.950637070278067e-06, "loss": 0.0012, "step": 25760 }, { "epoch": 3.72, "learning_rate": 3.940620242006572e-06, "loss": 0.0005, "step": 25765 }, { "epoch": 3.72, "learning_rate": 3.930603413735075e-06, "loss": 0.0008, "step": 25770 }, { "epoch": 3.72, "learning_rate": 3.920586585463579e-06, "loss": 0.0005, "step": 25775 }, { "epoch": 3.72, "learning_rate": 3.910569757192082e-06, "loss": 0.0007, "step": 25780 }, { "epoch": 3.72, "learning_rate": 3.900552928920587e-06, "loss": 0.001, "step": 25785 }, { "epoch": 3.72, "learning_rate": 3.890536100649091e-06, "loss": 0.0006, "step": 25790 }, { "epoch": 3.72, "learning_rate": 3.8805192723775946e-06, "loss": 0.0009, "step": 25795 }, { "epoch": 3.72, "learning_rate": 3.870502444106099e-06, "loss": 0.0004, "step": 25800 }, { "epoch": 3.72, "learning_rate": 3.8604856158346025e-06, "loss": 0.0006, "step": 25805 }, { "epoch": 3.72, "learning_rate": 3.850468787563107e-06, "loss": 0.0008, "step": 25810 }, { "epoch": 3.72, "learning_rate": 3.8404519592916095e-06, "loss": 0.0007, "step": 25815 }, { "epoch": 3.72, "learning_rate": 3.830435131020114e-06, "loss": 0.0004, "step": 25820 }, { "epoch": 3.72, "learning_rate": 3.8204183027486174e-06, "loss": 0.001, "step": 25825 }, { "epoch": 3.73, "learning_rate": 3.810401474477122e-06, "loss": 0.0005, "step": 25830 }, { "epoch": 3.73, "learning_rate": 3.8003846462056258e-06, "loss": 0.0008, "step": 25835 }, { "epoch": 3.73, "learning_rate": 3.7903678179341297e-06, "loss": 0.0008, "step": 25840 }, { "epoch": 3.73, "learning_rate": 3.7803509896626337e-06, "loss": 0.0007, "step": 25845 }, { "epoch": 3.73, "learning_rate": 3.7703341613911377e-06, "loss": 0.0009, "step": 25850 }, { "epoch": 3.73, "learning_rate": 3.7603173331196408e-06, "loss": 0.0007, "step": 25855 }, { "epoch": 3.73, "learning_rate": 3.7503005048481447e-06, "loss": 0.0006, "step": 25860 }, { "epoch": 3.73, "learning_rate": 3.7402836765766487e-06, "loss": 0.0008, "step": 25865 }, { "epoch": 3.73, "learning_rate": 3.7302668483051526e-06, "loss": 0.0006, "step": 25870 }, { "epoch": 3.73, "learning_rate": 3.720250020033657e-06, "loss": 0.001, "step": 25875 }, { "epoch": 3.73, "learning_rate": 3.710233191762161e-06, "loss": 0.0006, "step": 25880 }, { "epoch": 3.73, "learning_rate": 3.700216363490665e-06, "loss": 0.0005, "step": 25885 }, { "epoch": 3.73, "learning_rate": 3.690199535219169e-06, "loss": 0.001, "step": 25890 }, { "epoch": 3.74, "learning_rate": 3.680182706947672e-06, "loss": 0.001, "step": 25895 }, { "epoch": 3.74, "learning_rate": 3.670165878676176e-06, "loss": 0.0008, "step": 25900 }, { "epoch": 3.74, "learning_rate": 3.66014905040468e-06, "loss": 0.0005, "step": 25905 }, { "epoch": 3.74, "learning_rate": 3.650132222133184e-06, "loss": 0.0008, "step": 25910 }, { "epoch": 3.74, "learning_rate": 3.640115393861688e-06, "loss": 0.0006, "step": 25915 }, { "epoch": 3.74, "learning_rate": 3.6300985655901918e-06, "loss": 0.0007, "step": 25920 }, { "epoch": 3.74, "learning_rate": 3.6200817373186957e-06, "loss": 0.0006, "step": 25925 }, { "epoch": 3.74, "learning_rate": 3.6100649090471993e-06, "loss": 0.0009, "step": 25930 }, { "epoch": 3.74, "learning_rate": 3.6000480807757032e-06, "loss": 0.0006, "step": 25935 }, { "epoch": 3.74, "learning_rate": 3.590031252504207e-06, "loss": 0.0008, "step": 25940 }, { "epoch": 3.74, "learning_rate": 3.580014424232711e-06, "loss": 0.0009, "step": 25945 }, { "epoch": 3.74, "learning_rate": 3.569997595961215e-06, "loss": 0.001, "step": 25950 }, { "epoch": 3.74, "learning_rate": 3.559980767689719e-06, "loss": 0.0012, "step": 25955 }, { "epoch": 3.74, "learning_rate": 3.549963939418223e-06, "loss": 0.0009, "step": 25960 }, { "epoch": 3.75, "learning_rate": 3.539947111146727e-06, "loss": 0.0007, "step": 25965 }, { "epoch": 3.75, "learning_rate": 3.52993028287523e-06, "loss": 0.0007, "step": 25970 }, { "epoch": 3.75, "learning_rate": 3.519913454603734e-06, "loss": 0.0006, "step": 25975 }, { "epoch": 3.75, "learning_rate": 3.509896626332238e-06, "loss": 0.0007, "step": 25980 }, { "epoch": 3.75, "learning_rate": 3.4998797980607424e-06, "loss": 0.0006, "step": 25985 }, { "epoch": 3.75, "learning_rate": 3.4898629697892463e-06, "loss": 0.001, "step": 25990 }, { "epoch": 3.75, "learning_rate": 3.4798461415177503e-06, "loss": 0.0007, "step": 25995 }, { "epoch": 3.75, "learning_rate": 3.4698293132462542e-06, "loss": 0.0006, "step": 26000 }, { "epoch": 3.75, "learning_rate": 3.459812484974758e-06, "loss": 0.0009, "step": 26005 }, { "epoch": 3.75, "learning_rate": 3.4497956567032613e-06, "loss": 0.0006, "step": 26010 }, { "epoch": 3.75, "learning_rate": 3.4397788284317652e-06, "loss": 0.0006, "step": 26015 }, { "epoch": 3.75, "learning_rate": 3.429762000160269e-06, "loss": 0.0006, "step": 26020 }, { "epoch": 3.75, "learning_rate": 3.419745171888773e-06, "loss": 0.0005, "step": 26025 }, { "epoch": 3.75, "learning_rate": 3.409728343617277e-06, "loss": 0.0007, "step": 26030 }, { "epoch": 3.76, "learning_rate": 3.399711515345781e-06, "loss": 0.0009, "step": 26035 }, { "epoch": 3.76, "learning_rate": 3.389694687074285e-06, "loss": 0.0007, "step": 26040 }, { "epoch": 3.76, "learning_rate": 3.379677858802789e-06, "loss": 0.0007, "step": 26045 }, { "epoch": 3.76, "learning_rate": 3.3696610305312925e-06, "loss": 0.0008, "step": 26050 }, { "epoch": 3.76, "learning_rate": 3.3596442022597965e-06, "loss": 0.0007, "step": 26055 }, { "epoch": 3.76, "learning_rate": 3.3496273739883004e-06, "loss": 0.0006, "step": 26060 }, { "epoch": 3.76, "learning_rate": 3.3396105457168044e-06, "loss": 0.0007, "step": 26065 }, { "epoch": 3.76, "learning_rate": 3.3295937174453083e-06, "loss": 0.0006, "step": 26070 }, { "epoch": 3.76, "learning_rate": 3.3195768891738123e-06, "loss": 0.0007, "step": 26075 }, { "epoch": 3.76, "learning_rate": 3.3095600609023162e-06, "loss": 0.0008, "step": 26080 }, { "epoch": 3.76, "learning_rate": 3.29954323263082e-06, "loss": 0.0007, "step": 26085 }, { "epoch": 3.76, "learning_rate": 3.2895264043593233e-06, "loss": 0.001, "step": 26090 }, { "epoch": 3.76, "learning_rate": 3.2795095760878277e-06, "loss": 0.0008, "step": 26095 }, { "epoch": 3.76, "learning_rate": 3.2694927478163317e-06, "loss": 0.0006, "step": 26100 }, { "epoch": 3.77, "learning_rate": 3.2594759195448356e-06, "loss": 0.0007, "step": 26105 }, { "epoch": 3.77, "learning_rate": 3.2494590912733396e-06, "loss": 0.0008, "step": 26110 }, { "epoch": 3.77, "learning_rate": 3.2394422630018435e-06, "loss": 0.0009, "step": 26115 }, { "epoch": 3.77, "learning_rate": 3.2294254347303475e-06, "loss": 0.0004, "step": 26120 }, { "epoch": 3.77, "learning_rate": 3.2194086064588514e-06, "loss": 0.0008, "step": 26125 }, { "epoch": 3.77, "learning_rate": 3.2093917781873545e-06, "loss": 0.0011, "step": 26130 }, { "epoch": 3.77, "learning_rate": 3.1993749499158585e-06, "loss": 0.0007, "step": 26135 }, { "epoch": 3.77, "learning_rate": 3.1893581216443625e-06, "loss": 0.0007, "step": 26140 }, { "epoch": 3.77, "learning_rate": 3.1793412933728664e-06, "loss": 0.001, "step": 26145 }, { "epoch": 3.77, "learning_rate": 3.1693244651013704e-06, "loss": 0.0004, "step": 26150 }, { "epoch": 3.77, "learning_rate": 3.1593076368298743e-06, "loss": 0.0004, "step": 26155 }, { "epoch": 3.77, "learning_rate": 3.1492908085583783e-06, "loss": 0.0007, "step": 26160 }, { "epoch": 3.77, "learning_rate": 3.1392739802868822e-06, "loss": 0.0006, "step": 26165 }, { "epoch": 3.77, "learning_rate": 3.1292571520153858e-06, "loss": 0.0007, "step": 26170 }, { "epoch": 3.78, "learning_rate": 3.11924032374389e-06, "loss": 0.0008, "step": 26175 }, { "epoch": 3.78, "learning_rate": 3.1092234954723937e-06, "loss": 0.0004, "step": 26180 }, { "epoch": 3.78, "learning_rate": 3.0992066672008976e-06, "loss": 0.0008, "step": 26185 }, { "epoch": 3.78, "learning_rate": 3.0891898389294016e-06, "loss": 0.0005, "step": 26190 }, { "epoch": 3.78, "learning_rate": 3.0791730106579055e-06, "loss": 0.0009, "step": 26195 }, { "epoch": 3.78, "learning_rate": 3.069156182386409e-06, "loss": 0.001, "step": 26200 }, { "epoch": 3.78, "learning_rate": 3.059139354114913e-06, "loss": 0.0008, "step": 26205 }, { "epoch": 3.78, "learning_rate": 3.049122525843417e-06, "loss": 0.0005, "step": 26210 }, { "epoch": 3.78, "learning_rate": 3.039105697571921e-06, "loss": 0.0006, "step": 26215 }, { "epoch": 3.78, "learning_rate": 3.029088869300425e-06, "loss": 0.0006, "step": 26220 }, { "epoch": 3.78, "learning_rate": 3.019072041028929e-06, "loss": 0.0009, "step": 26225 }, { "epoch": 3.78, "learning_rate": 3.009055212757433e-06, "loss": 0.0009, "step": 26230 }, { "epoch": 3.78, "learning_rate": 2.9990383844859368e-06, "loss": 0.0006, "step": 26235 }, { "epoch": 3.78, "learning_rate": 2.9890215562144403e-06, "loss": 0.0006, "step": 26240 }, { "epoch": 3.79, "learning_rate": 2.9790047279429443e-06, "loss": 0.0008, "step": 26245 }, { "epoch": 3.79, "learning_rate": 2.9689878996714482e-06, "loss": 0.0009, "step": 26250 }, { "epoch": 3.79, "learning_rate": 2.9589710713999518e-06, "loss": 0.0006, "step": 26255 }, { "epoch": 3.79, "learning_rate": 2.9489542431284557e-06, "loss": 0.0006, "step": 26260 }, { "epoch": 3.79, "learning_rate": 2.9389374148569597e-06, "loss": 0.0009, "step": 26265 }, { "epoch": 3.79, "learning_rate": 2.9289205865854636e-06, "loss": 0.001, "step": 26270 }, { "epoch": 3.79, "learning_rate": 2.9189037583139676e-06, "loss": 0.0006, "step": 26275 }, { "epoch": 3.79, "learning_rate": 2.9088869300424715e-06, "loss": 0.0005, "step": 26280 }, { "epoch": 3.79, "learning_rate": 2.8988701017709755e-06, "loss": 0.0007, "step": 26285 }, { "epoch": 3.79, "learning_rate": 2.8888532734994794e-06, "loss": 0.0009, "step": 26290 }, { "epoch": 3.79, "learning_rate": 2.878836445227983e-06, "loss": 0.0005, "step": 26295 }, { "epoch": 3.79, "learning_rate": 2.868819616956487e-06, "loss": 0.0006, "step": 26300 }, { "epoch": 3.79, "learning_rate": 2.858802788684991e-06, "loss": 0.0007, "step": 26305 }, { "epoch": 3.79, "learning_rate": 2.848785960413495e-06, "loss": 0.0007, "step": 26310 }, { "epoch": 3.8, "learning_rate": 2.8387691321419984e-06, "loss": 0.0008, "step": 26315 }, { "epoch": 3.8, "learning_rate": 2.8287523038705023e-06, "loss": 0.0005, "step": 26320 }, { "epoch": 3.8, "learning_rate": 2.8187354755990063e-06, "loss": 0.0009, "step": 26325 }, { "epoch": 3.8, "learning_rate": 2.8087186473275102e-06, "loss": 0.0006, "step": 26330 }, { "epoch": 3.8, "learning_rate": 2.798701819056014e-06, "loss": 0.0009, "step": 26335 }, { "epoch": 3.8, "learning_rate": 2.788684990784518e-06, "loss": 0.0007, "step": 26340 }, { "epoch": 3.8, "learning_rate": 2.778668162513022e-06, "loss": 0.0008, "step": 26345 }, { "epoch": 3.8, "learning_rate": 2.768651334241526e-06, "loss": 0.0005, "step": 26350 }, { "epoch": 3.8, "learning_rate": 2.7586345059700296e-06, "loss": 0.0008, "step": 26355 }, { "epoch": 3.8, "learning_rate": 2.7486176776985336e-06, "loss": 0.0004, "step": 26360 }, { "epoch": 3.8, "learning_rate": 2.7386008494270375e-06, "loss": 0.0008, "step": 26365 }, { "epoch": 3.8, "learning_rate": 2.7285840211555415e-06, "loss": 0.0009, "step": 26370 }, { "epoch": 3.8, "learning_rate": 2.718567192884045e-06, "loss": 0.0006, "step": 26375 }, { "epoch": 3.8, "learning_rate": 2.708550364612549e-06, "loss": 0.0007, "step": 26380 }, { "epoch": 3.81, "learning_rate": 2.6985335363410533e-06, "loss": 0.0006, "step": 26385 }, { "epoch": 3.81, "learning_rate": 2.6885167080695573e-06, "loss": 0.0011, "step": 26390 }, { "epoch": 3.81, "learning_rate": 2.678499879798061e-06, "loss": 0.0005, "step": 26395 }, { "epoch": 3.81, "learning_rate": 2.6684830515265648e-06, "loss": 0.0007, "step": 26400 }, { "epoch": 3.81, "learning_rate": 2.6584662232550687e-06, "loss": 0.0006, "step": 26405 }, { "epoch": 3.81, "learning_rate": 2.6484493949835727e-06, "loss": 0.0005, "step": 26410 }, { "epoch": 3.81, "learning_rate": 2.6384325667120762e-06, "loss": 0.0007, "step": 26415 }, { "epoch": 3.81, "learning_rate": 2.62841573844058e-06, "loss": 0.0008, "step": 26420 }, { "epoch": 3.81, "learning_rate": 2.618398910169084e-06, "loss": 0.0008, "step": 26425 }, { "epoch": 3.81, "learning_rate": 2.608382081897588e-06, "loss": 0.0006, "step": 26430 }, { "epoch": 3.81, "learning_rate": 2.5983652536260916e-06, "loss": 0.001, "step": 26435 }, { "epoch": 3.81, "learning_rate": 2.588348425354596e-06, "loss": 0.0011, "step": 26440 }, { "epoch": 3.81, "learning_rate": 2.5783315970831e-06, "loss": 0.0006, "step": 26445 }, { "epoch": 3.82, "learning_rate": 2.568314768811604e-06, "loss": 0.001, "step": 26450 }, { "epoch": 3.82, "learning_rate": 2.5582979405401075e-06, "loss": 0.0007, "step": 26455 }, { "epoch": 3.82, "learning_rate": 2.5482811122686114e-06, "loss": 0.001, "step": 26460 }, { "epoch": 3.82, "learning_rate": 2.5382642839971154e-06, "loss": 0.0005, "step": 26465 }, { "epoch": 3.82, "learning_rate": 2.5282474557256193e-06, "loss": 0.0005, "step": 26470 }, { "epoch": 3.82, "learning_rate": 2.518230627454123e-06, "loss": 0.0005, "step": 26475 }, { "epoch": 3.82, "learning_rate": 2.508213799182627e-06, "loss": 0.0005, "step": 26480 }, { "epoch": 3.82, "learning_rate": 2.4981969709111308e-06, "loss": 0.0007, "step": 26485 }, { "epoch": 3.82, "learning_rate": 2.4881801426396347e-06, "loss": 0.0006, "step": 26490 }, { "epoch": 3.82, "learning_rate": 2.4781633143681387e-06, "loss": 0.0008, "step": 26495 }, { "epoch": 3.82, "learning_rate": 2.4681464860966426e-06, "loss": 0.0011, "step": 26500 }, { "epoch": 3.82, "learning_rate": 2.4581296578251466e-06, "loss": 0.0008, "step": 26505 }, { "epoch": 3.82, "learning_rate": 2.4481128295536506e-06, "loss": 0.0007, "step": 26510 }, { "epoch": 3.82, "learning_rate": 2.438096001282154e-06, "loss": 0.0006, "step": 26515 }, { "epoch": 3.83, "learning_rate": 2.428079173010658e-06, "loss": 0.0007, "step": 26520 }, { "epoch": 3.83, "learning_rate": 2.418062344739162e-06, "loss": 0.001, "step": 26525 }, { "epoch": 3.83, "learning_rate": 2.408045516467666e-06, "loss": 0.0009, "step": 26530 }, { "epoch": 3.83, "learning_rate": 2.3980286881961695e-06, "loss": 0.0008, "step": 26535 }, { "epoch": 3.83, "learning_rate": 2.3880118599246734e-06, "loss": 0.0012, "step": 26540 }, { "epoch": 3.83, "learning_rate": 2.3779950316531774e-06, "loss": 0.0007, "step": 26545 }, { "epoch": 3.83, "learning_rate": 2.3679782033816814e-06, "loss": 0.0009, "step": 26550 }, { "epoch": 3.83, "learning_rate": 2.3579613751101853e-06, "loss": 0.0005, "step": 26555 }, { "epoch": 3.83, "learning_rate": 2.3479445468386893e-06, "loss": 0.0009, "step": 26560 }, { "epoch": 3.83, "learning_rate": 2.3379277185671932e-06, "loss": 0.0004, "step": 26565 }, { "epoch": 3.83, "learning_rate": 2.327910890295697e-06, "loss": 0.0006, "step": 26570 }, { "epoch": 3.83, "learning_rate": 2.3178940620242007e-06, "loss": 0.0004, "step": 26575 }, { "epoch": 3.83, "learning_rate": 2.3078772337527047e-06, "loss": 0.0009, "step": 26580 }, { "epoch": 3.83, "learning_rate": 2.2978604054812086e-06, "loss": 0.0009, "step": 26585 }, { "epoch": 3.84, "learning_rate": 2.2878435772097126e-06, "loss": 0.0008, "step": 26590 }, { "epoch": 3.84, "learning_rate": 2.277826748938216e-06, "loss": 0.0008, "step": 26595 }, { "epoch": 3.84, "learning_rate": 2.26780992066672e-06, "loss": 0.0009, "step": 26600 }, { "epoch": 3.84, "learning_rate": 2.257793092395224e-06, "loss": 0.0004, "step": 26605 }, { "epoch": 3.84, "learning_rate": 2.247776264123728e-06, "loss": 0.0008, "step": 26610 }, { "epoch": 3.84, "learning_rate": 2.237759435852232e-06, "loss": 0.0006, "step": 26615 }, { "epoch": 3.84, "learning_rate": 2.227742607580736e-06, "loss": 0.0009, "step": 26620 }, { "epoch": 3.84, "learning_rate": 2.21772577930924e-06, "loss": 0.0007, "step": 26625 }, { "epoch": 3.84, "learning_rate": 2.2077089510377434e-06, "loss": 0.0008, "step": 26630 }, { "epoch": 3.84, "learning_rate": 2.1976921227662473e-06, "loss": 0.0008, "step": 26635 }, { "epoch": 3.84, "learning_rate": 2.1876752944947513e-06, "loss": 0.0007, "step": 26640 }, { "epoch": 3.84, "learning_rate": 2.1776584662232553e-06, "loss": 0.001, "step": 26645 }, { "epoch": 3.84, "learning_rate": 2.1676416379517588e-06, "loss": 0.0008, "step": 26650 }, { "epoch": 3.84, "learning_rate": 2.1576248096802627e-06, "loss": 0.0006, "step": 26655 }, { "epoch": 3.85, "learning_rate": 2.1476079814087667e-06, "loss": 0.0006, "step": 26660 }, { "epoch": 3.85, "learning_rate": 2.1375911531372707e-06, "loss": 0.0007, "step": 26665 }, { "epoch": 3.85, "learning_rate": 2.1275743248657746e-06, "loss": 0.0005, "step": 26670 }, { "epoch": 3.85, "learning_rate": 2.1175574965942786e-06, "loss": 0.0007, "step": 26675 }, { "epoch": 3.85, "learning_rate": 2.1075406683227825e-06, "loss": 0.0007, "step": 26680 }, { "epoch": 3.85, "learning_rate": 2.0975238400512865e-06, "loss": 0.0006, "step": 26685 }, { "epoch": 3.85, "learning_rate": 2.08750701177979e-06, "loss": 0.0009, "step": 26690 }, { "epoch": 3.85, "learning_rate": 2.077490183508294e-06, "loss": 0.0007, "step": 26695 }, { "epoch": 3.85, "learning_rate": 2.067473355236798e-06, "loss": 0.0009, "step": 26700 }, { "epoch": 3.85, "learning_rate": 2.057456526965302e-06, "loss": 0.0007, "step": 26705 }, { "epoch": 3.85, "learning_rate": 2.0474396986938054e-06, "loss": 0.0008, "step": 26710 }, { "epoch": 3.85, "learning_rate": 2.0374228704223094e-06, "loss": 0.0007, "step": 26715 }, { "epoch": 3.85, "learning_rate": 2.0274060421508133e-06, "loss": 0.0006, "step": 26720 }, { "epoch": 3.85, "learning_rate": 2.0173892138793173e-06, "loss": 0.001, "step": 26725 }, { "epoch": 3.86, "learning_rate": 2.0073723856078212e-06, "loss": 0.0008, "step": 26730 }, { "epoch": 3.86, "learning_rate": 1.997355557336325e-06, "loss": 0.0008, "step": 26735 }, { "epoch": 3.86, "learning_rate": 1.987338729064829e-06, "loss": 0.0008, "step": 26740 }, { "epoch": 3.86, "learning_rate": 1.977321900793333e-06, "loss": 0.0007, "step": 26745 }, { "epoch": 3.86, "learning_rate": 1.9673050725218366e-06, "loss": 0.0006, "step": 26750 }, { "epoch": 3.86, "learning_rate": 1.9572882442503406e-06, "loss": 0.0005, "step": 26755 }, { "epoch": 3.86, "learning_rate": 1.9472714159788446e-06, "loss": 0.0005, "step": 26760 }, { "epoch": 3.86, "learning_rate": 1.9372545877073485e-06, "loss": 0.001, "step": 26765 }, { "epoch": 3.86, "learning_rate": 1.927237759435852e-06, "loss": 0.0006, "step": 26770 }, { "epoch": 3.86, "learning_rate": 1.917220931164356e-06, "loss": 0.0007, "step": 26775 }, { "epoch": 3.86, "learning_rate": 1.9072041028928602e-06, "loss": 0.0006, "step": 26780 }, { "epoch": 3.86, "learning_rate": 1.8971872746213641e-06, "loss": 0.0007, "step": 26785 }, { "epoch": 3.86, "learning_rate": 1.8871704463498677e-06, "loss": 0.0008, "step": 26790 }, { "epoch": 3.86, "learning_rate": 1.8771536180783716e-06, "loss": 0.0007, "step": 26795 }, { "epoch": 3.87, "learning_rate": 1.8671367898068758e-06, "loss": 0.0007, "step": 26800 }, { "epoch": 3.87, "learning_rate": 1.8571199615353797e-06, "loss": 0.0008, "step": 26805 }, { "epoch": 3.87, "learning_rate": 1.8471031332638833e-06, "loss": 0.0008, "step": 26810 }, { "epoch": 3.87, "learning_rate": 1.8370863049923872e-06, "loss": 0.0008, "step": 26815 }, { "epoch": 3.87, "learning_rate": 1.8270694767208912e-06, "loss": 0.001, "step": 26820 }, { "epoch": 3.87, "learning_rate": 1.8170526484493951e-06, "loss": 0.001, "step": 26825 }, { "epoch": 3.87, "learning_rate": 1.8070358201778989e-06, "loss": 0.0007, "step": 26830 }, { "epoch": 3.87, "learning_rate": 1.7970189919064028e-06, "loss": 0.0006, "step": 26835 }, { "epoch": 3.87, "learning_rate": 1.7870021636349068e-06, "loss": 0.001, "step": 26840 }, { "epoch": 3.87, "learning_rate": 1.7769853353634107e-06, "loss": 0.0009, "step": 26845 }, { "epoch": 3.87, "learning_rate": 1.7669685070919145e-06, "loss": 0.001, "step": 26850 }, { "epoch": 3.87, "learning_rate": 1.7569516788204184e-06, "loss": 0.0003, "step": 26855 }, { "epoch": 3.87, "learning_rate": 1.7469348505489224e-06, "loss": 0.0008, "step": 26860 }, { "epoch": 3.87, "learning_rate": 1.7369180222774264e-06, "loss": 0.0004, "step": 26865 }, { "epoch": 3.88, "learning_rate": 1.72690119400593e-06, "loss": 0.001, "step": 26870 }, { "epoch": 3.88, "learning_rate": 1.7168843657344339e-06, "loss": 0.0008, "step": 26875 }, { "epoch": 3.88, "learning_rate": 1.7068675374629378e-06, "loss": 0.0007, "step": 26880 }, { "epoch": 3.88, "learning_rate": 1.6968507091914418e-06, "loss": 0.0007, "step": 26885 }, { "epoch": 3.88, "learning_rate": 1.6868338809199455e-06, "loss": 0.0008, "step": 26890 }, { "epoch": 3.88, "learning_rate": 1.6768170526484495e-06, "loss": 0.0007, "step": 26895 }, { "epoch": 3.88, "learning_rate": 1.6668002243769534e-06, "loss": 0.0007, "step": 26900 }, { "epoch": 3.88, "learning_rate": 1.6567833961054574e-06, "loss": 0.0006, "step": 26905 }, { "epoch": 3.88, "learning_rate": 1.6467665678339611e-06, "loss": 0.0005, "step": 26910 }, { "epoch": 3.88, "learning_rate": 1.636749739562465e-06, "loss": 0.0005, "step": 26915 }, { "epoch": 3.88, "learning_rate": 1.626732911290969e-06, "loss": 0.0008, "step": 26920 }, { "epoch": 3.88, "learning_rate": 1.616716083019473e-06, "loss": 0.0009, "step": 26925 }, { "epoch": 3.88, "learning_rate": 1.6066992547479765e-06, "loss": 0.001, "step": 26930 }, { "epoch": 3.89, "learning_rate": 1.5966824264764805e-06, "loss": 0.0008, "step": 26935 }, { "epoch": 3.89, "learning_rate": 1.5866655982049844e-06, "loss": 0.0005, "step": 26940 }, { "epoch": 3.89, "learning_rate": 1.5766487699334886e-06, "loss": 0.0005, "step": 26945 }, { "epoch": 3.89, "learning_rate": 1.5666319416619921e-06, "loss": 0.0004, "step": 26950 }, { "epoch": 3.89, "learning_rate": 1.556615113390496e-06, "loss": 0.0006, "step": 26955 }, { "epoch": 3.89, "learning_rate": 1.5465982851189998e-06, "loss": 0.0009, "step": 26960 }, { "epoch": 3.89, "learning_rate": 1.5365814568475038e-06, "loss": 0.0009, "step": 26965 }, { "epoch": 3.89, "learning_rate": 1.5265646285760077e-06, "loss": 0.0008, "step": 26970 }, { "epoch": 3.89, "learning_rate": 1.5165478003045117e-06, "loss": 0.0007, "step": 26975 }, { "epoch": 3.89, "learning_rate": 1.5065309720330154e-06, "loss": 0.0007, "step": 26980 }, { "epoch": 3.89, "learning_rate": 1.4965141437615194e-06, "loss": 0.0005, "step": 26985 }, { "epoch": 3.89, "learning_rate": 1.4864973154900231e-06, "loss": 0.0006, "step": 26990 }, { "epoch": 3.89, "learning_rate": 1.476480487218527e-06, "loss": 0.0009, "step": 26995 }, { "epoch": 3.89, "learning_rate": 1.466463658947031e-06, "loss": 0.0007, "step": 27000 }, { "epoch": 3.9, "learning_rate": 1.456446830675535e-06, "loss": 0.001, "step": 27005 }, { "epoch": 3.9, "learning_rate": 1.4464300024040388e-06, "loss": 0.0009, "step": 27010 }, { "epoch": 3.9, "learning_rate": 1.4364131741325427e-06, "loss": 0.0009, "step": 27015 }, { "epoch": 3.9, "learning_rate": 1.4263963458610465e-06, "loss": 0.0007, "step": 27020 }, { "epoch": 3.9, "learning_rate": 1.4163795175895506e-06, "loss": 0.0007, "step": 27025 }, { "epoch": 3.9, "learning_rate": 1.4063626893180544e-06, "loss": 0.0008, "step": 27030 }, { "epoch": 3.9, "learning_rate": 1.3963458610465583e-06, "loss": 0.0004, "step": 27035 }, { "epoch": 3.9, "learning_rate": 1.386329032775062e-06, "loss": 0.0004, "step": 27040 }, { "epoch": 3.9, "learning_rate": 1.376312204503566e-06, "loss": 0.0008, "step": 27045 }, { "epoch": 3.9, "learning_rate": 1.3662953762320698e-06, "loss": 0.001, "step": 27050 }, { "epoch": 3.9, "learning_rate": 1.356278547960574e-06, "loss": 0.0004, "step": 27055 }, { "epoch": 3.9, "learning_rate": 1.3462617196890777e-06, "loss": 0.0005, "step": 27060 }, { "epoch": 3.9, "learning_rate": 1.3362448914175816e-06, "loss": 0.0009, "step": 27065 }, { "epoch": 3.9, "learning_rate": 1.3262280631460854e-06, "loss": 0.0007, "step": 27070 }, { "epoch": 3.91, "learning_rate": 1.3162112348745893e-06, "loss": 0.0007, "step": 27075 }, { "epoch": 3.91, "learning_rate": 1.3061944066030933e-06, "loss": 0.0012, "step": 27080 }, { "epoch": 3.91, "learning_rate": 1.2961775783315973e-06, "loss": 0.0007, "step": 27085 }, { "epoch": 3.91, "learning_rate": 1.286160750060101e-06, "loss": 0.0009, "step": 27090 }, { "epoch": 3.91, "learning_rate": 1.276143921788605e-06, "loss": 0.0007, "step": 27095 }, { "epoch": 3.91, "learning_rate": 1.2661270935171087e-06, "loss": 0.0011, "step": 27100 }, { "epoch": 3.91, "learning_rate": 1.2561102652456127e-06, "loss": 0.001, "step": 27105 }, { "epoch": 3.91, "learning_rate": 1.2460934369741166e-06, "loss": 0.0006, "step": 27110 }, { "epoch": 3.91, "learning_rate": 1.2360766087026206e-06, "loss": 0.0006, "step": 27115 }, { "epoch": 3.91, "learning_rate": 1.2260597804311243e-06, "loss": 0.0007, "step": 27120 }, { "epoch": 3.91, "learning_rate": 1.2160429521596283e-06, "loss": 0.0005, "step": 27125 }, { "epoch": 3.91, "learning_rate": 1.206026123888132e-06, "loss": 0.0004, "step": 27130 }, { "epoch": 3.91, "learning_rate": 1.196009295616636e-06, "loss": 0.0005, "step": 27135 }, { "epoch": 3.91, "learning_rate": 1.18599246734514e-06, "loss": 0.0009, "step": 27140 }, { "epoch": 3.92, "learning_rate": 1.1759756390736439e-06, "loss": 0.0007, "step": 27145 }, { "epoch": 3.92, "learning_rate": 1.1659588108021476e-06, "loss": 0.0005, "step": 27150 }, { "epoch": 3.92, "learning_rate": 1.1559419825306516e-06, "loss": 0.0005, "step": 27155 }, { "epoch": 3.92, "learning_rate": 1.1459251542591553e-06, "loss": 0.0008, "step": 27160 }, { "epoch": 3.92, "learning_rate": 1.1359083259876593e-06, "loss": 0.0006, "step": 27165 }, { "epoch": 3.92, "learning_rate": 1.1258914977161632e-06, "loss": 0.0006, "step": 27170 }, { "epoch": 3.92, "learning_rate": 1.1158746694446672e-06, "loss": 0.0006, "step": 27175 }, { "epoch": 3.92, "learning_rate": 1.105857841173171e-06, "loss": 0.0007, "step": 27180 }, { "epoch": 3.92, "learning_rate": 1.095841012901675e-06, "loss": 0.0008, "step": 27185 }, { "epoch": 3.92, "learning_rate": 1.0858241846301786e-06, "loss": 0.0008, "step": 27190 }, { "epoch": 3.92, "learning_rate": 1.0758073563586826e-06, "loss": 0.0009, "step": 27195 }, { "epoch": 3.92, "learning_rate": 1.0657905280871866e-06, "loss": 0.0007, "step": 27200 }, { "epoch": 3.92, "learning_rate": 1.0557736998156905e-06, "loss": 0.001, "step": 27205 }, { "epoch": 3.92, "learning_rate": 1.0457568715441943e-06, "loss": 0.0009, "step": 27210 }, { "epoch": 3.93, "learning_rate": 1.0357400432726982e-06, "loss": 0.0008, "step": 27215 }, { "epoch": 3.93, "learning_rate": 1.025723215001202e-06, "loss": 0.0009, "step": 27220 }, { "epoch": 3.93, "learning_rate": 1.0157063867297061e-06, "loss": 0.0008, "step": 27225 }, { "epoch": 3.93, "learning_rate": 1.0056895584582099e-06, "loss": 0.0007, "step": 27230 }, { "epoch": 3.93, "learning_rate": 9.956727301867138e-07, "loss": 0.0007, "step": 27235 }, { "epoch": 3.93, "learning_rate": 9.856559019152176e-07, "loss": 0.0008, "step": 27240 }, { "epoch": 3.93, "learning_rate": 9.756390736437215e-07, "loss": 0.0008, "step": 27245 }, { "epoch": 3.93, "learning_rate": 9.656222453722253e-07, "loss": 0.0007, "step": 27250 }, { "epoch": 3.93, "learning_rate": 9.556054171007294e-07, "loss": 0.0008, "step": 27255 }, { "epoch": 3.93, "learning_rate": 9.455885888292332e-07, "loss": 0.0008, "step": 27260 }, { "epoch": 3.93, "learning_rate": 9.355717605577371e-07, "loss": 0.0007, "step": 27265 }, { "epoch": 3.93, "learning_rate": 9.255549322862409e-07, "loss": 0.0011, "step": 27270 }, { "epoch": 3.93, "learning_rate": 9.155381040147448e-07, "loss": 0.0006, "step": 27275 }, { "epoch": 3.93, "learning_rate": 9.055212757432487e-07, "loss": 0.0009, "step": 27280 }, { "epoch": 3.94, "learning_rate": 8.955044474717526e-07, "loss": 0.0008, "step": 27285 }, { "epoch": 3.94, "learning_rate": 8.854876192002565e-07, "loss": 0.0008, "step": 27290 }, { "epoch": 3.94, "learning_rate": 8.754707909287605e-07, "loss": 0.0008, "step": 27295 }, { "epoch": 3.94, "learning_rate": 8.654539626572642e-07, "loss": 0.0008, "step": 27300 }, { "epoch": 3.94, "learning_rate": 8.554371343857683e-07, "loss": 0.0005, "step": 27305 }, { "epoch": 3.94, "learning_rate": 8.45420306114272e-07, "loss": 0.0005, "step": 27310 }, { "epoch": 3.94, "learning_rate": 8.354034778427759e-07, "loss": 0.0009, "step": 27315 }, { "epoch": 3.94, "learning_rate": 8.253866495712798e-07, "loss": 0.0007, "step": 27320 }, { "epoch": 3.94, "learning_rate": 8.153698212997836e-07, "loss": 0.0008, "step": 27325 }, { "epoch": 3.94, "learning_rate": 8.053529930282876e-07, "loss": 0.0007, "step": 27330 }, { "epoch": 3.94, "learning_rate": 7.953361647567914e-07, "loss": 0.0008, "step": 27335 }, { "epoch": 3.94, "learning_rate": 7.853193364852953e-07, "loss": 0.0006, "step": 27340 }, { "epoch": 3.94, "learning_rate": 7.753025082137993e-07, "loss": 0.0005, "step": 27345 }, { "epoch": 3.94, "learning_rate": 7.652856799423031e-07, "loss": 0.0009, "step": 27350 }, { "epoch": 3.95, "learning_rate": 7.55268851670807e-07, "loss": 0.0011, "step": 27355 }, { "epoch": 3.95, "learning_rate": 7.452520233993109e-07, "loss": 0.0008, "step": 27360 }, { "epoch": 3.95, "learning_rate": 7.352351951278148e-07, "loss": 0.0005, "step": 27365 }, { "epoch": 3.95, "learning_rate": 7.252183668563186e-07, "loss": 0.001, "step": 27370 }, { "epoch": 3.95, "learning_rate": 7.152015385848226e-07, "loss": 0.0008, "step": 27375 }, { "epoch": 3.95, "learning_rate": 7.051847103133264e-07, "loss": 0.0007, "step": 27380 }, { "epoch": 3.95, "learning_rate": 6.951678820418303e-07, "loss": 0.0009, "step": 27385 }, { "epoch": 3.95, "learning_rate": 6.851510537703342e-07, "loss": 0.0009, "step": 27390 }, { "epoch": 3.95, "learning_rate": 6.751342254988381e-07, "loss": 0.0007, "step": 27395 }, { "epoch": 3.95, "learning_rate": 6.651173972273419e-07, "loss": 0.0011, "step": 27400 }, { "epoch": 3.95, "learning_rate": 6.551005689558459e-07, "loss": 0.0008, "step": 27405 }, { "epoch": 3.95, "learning_rate": 6.450837406843498e-07, "loss": 0.0008, "step": 27410 }, { "epoch": 3.95, "learning_rate": 6.350669124128536e-07, "loss": 0.0007, "step": 27415 }, { "epoch": 3.95, "learning_rate": 6.250500841413576e-07, "loss": 0.0006, "step": 27420 }, { "epoch": 3.96, "learning_rate": 6.150332558698614e-07, "loss": 0.0008, "step": 27425 }, { "epoch": 3.96, "learning_rate": 6.050164275983653e-07, "loss": 0.0005, "step": 27430 }, { "epoch": 3.96, "learning_rate": 5.949995993268692e-07, "loss": 0.0007, "step": 27435 }, { "epoch": 3.96, "learning_rate": 5.849827710553731e-07, "loss": 0.0005, "step": 27440 }, { "epoch": 3.96, "learning_rate": 5.74965942783877e-07, "loss": 0.0007, "step": 27445 }, { "epoch": 3.96, "learning_rate": 5.649491145123809e-07, "loss": 0.0008, "step": 27450 }, { "epoch": 3.96, "learning_rate": 5.549322862408847e-07, "loss": 0.0008, "step": 27455 }, { "epoch": 3.96, "learning_rate": 5.449154579693887e-07, "loss": 0.0008, "step": 27460 }, { "epoch": 3.96, "learning_rate": 5.348986296978925e-07, "loss": 0.0009, "step": 27465 }, { "epoch": 3.96, "learning_rate": 5.248818014263964e-07, "loss": 0.0007, "step": 27470 }, { "epoch": 3.96, "learning_rate": 5.148649731549003e-07, "loss": 0.0008, "step": 27475 }, { "epoch": 3.96, "learning_rate": 5.048481448834042e-07, "loss": 0.001, "step": 27480 }, { "epoch": 3.96, "learning_rate": 4.94831316611908e-07, "loss": 0.0007, "step": 27485 }, { "epoch": 3.97, "learning_rate": 4.848144883404119e-07, "loss": 0.0006, "step": 27490 }, { "epoch": 3.97, "learning_rate": 4.7479766006891574e-07, "loss": 0.0005, "step": 27495 }, { "epoch": 3.97, "learning_rate": 4.6478083179741964e-07, "loss": 0.001, "step": 27500 }, { "epoch": 3.97, "learning_rate": 4.5476400352592354e-07, "loss": 0.0005, "step": 27505 }, { "epoch": 3.97, "learning_rate": 4.4474717525442745e-07, "loss": 0.0004, "step": 27510 }, { "epoch": 3.97, "learning_rate": 4.347303469829313e-07, "loss": 0.001, "step": 27515 }, { "epoch": 3.97, "learning_rate": 4.247135187114352e-07, "loss": 0.0006, "step": 27520 }, { "epoch": 3.97, "learning_rate": 4.146966904399391e-07, "loss": 0.0005, "step": 27525 }, { "epoch": 3.97, "learning_rate": 4.0467986216844295e-07, "loss": 0.001, "step": 27530 }, { "epoch": 3.97, "learning_rate": 3.9466303389694686e-07, "loss": 0.0006, "step": 27535 }, { "epoch": 3.97, "learning_rate": 3.846462056254508e-07, "loss": 0.0008, "step": 27540 }, { "epoch": 3.97, "learning_rate": 3.7462937735395466e-07, "loss": 0.0008, "step": 27545 }, { "epoch": 3.97, "learning_rate": 3.6461254908245857e-07, "loss": 0.001, "step": 27550 }, { "epoch": 3.97, "learning_rate": 3.5459572081096247e-07, "loss": 0.0004, "step": 27555 }, { "epoch": 3.98, "learning_rate": 3.445788925394663e-07, "loss": 0.0007, "step": 27560 }, { "epoch": 3.98, "learning_rate": 3.345620642679702e-07, "loss": 0.0004, "step": 27565 }, { "epoch": 3.98, "learning_rate": 3.2454523599647413e-07, "loss": 0.0007, "step": 27570 }, { "epoch": 3.98, "learning_rate": 3.14528407724978e-07, "loss": 0.0005, "step": 27575 }, { "epoch": 3.98, "learning_rate": 3.0451157945348183e-07, "loss": 0.0006, "step": 27580 }, { "epoch": 3.98, "learning_rate": 2.9449475118198573e-07, "loss": 0.0006, "step": 27585 }, { "epoch": 3.98, "learning_rate": 2.8447792291048963e-07, "loss": 0.0007, "step": 27590 }, { "epoch": 3.98, "learning_rate": 2.744610946389935e-07, "loss": 0.001, "step": 27595 }, { "epoch": 3.98, "learning_rate": 2.644442663674974e-07, "loss": 0.0007, "step": 27600 }, { "epoch": 3.98, "learning_rate": 2.544274380960013e-07, "loss": 0.0003, "step": 27605 }, { "epoch": 3.98, "learning_rate": 2.4441060982450514e-07, "loss": 0.0008, "step": 27610 }, { "epoch": 3.98, "learning_rate": 2.3439378155300904e-07, "loss": 0.0005, "step": 27615 }, { "epoch": 3.98, "learning_rate": 2.2437695328151295e-07, "loss": 0.0008, "step": 27620 }, { "epoch": 3.98, "learning_rate": 2.1436012501001682e-07, "loss": 0.0008, "step": 27625 }, { "epoch": 3.99, "learning_rate": 2.0434329673852073e-07, "loss": 0.0008, "step": 27630 }, { "epoch": 3.99, "learning_rate": 1.943264684670246e-07, "loss": 0.0008, "step": 27635 }, { "epoch": 3.99, "learning_rate": 1.8430964019552848e-07, "loss": 0.001, "step": 27640 }, { "epoch": 3.99, "learning_rate": 1.7429281192403239e-07, "loss": 0.0006, "step": 27645 }, { "epoch": 3.99, "learning_rate": 1.6427598365253626e-07, "loss": 0.0007, "step": 27650 }, { "epoch": 3.99, "learning_rate": 1.5425915538104017e-07, "loss": 0.0003, "step": 27655 }, { "epoch": 3.99, "learning_rate": 1.4424232710954404e-07, "loss": 0.001, "step": 27660 }, { "epoch": 3.99, "learning_rate": 1.3422549883804792e-07, "loss": 0.0008, "step": 27665 }, { "epoch": 3.99, "learning_rate": 1.2420867056655182e-07, "loss": 0.0007, "step": 27670 }, { "epoch": 3.99, "learning_rate": 1.141918422950557e-07, "loss": 0.0008, "step": 27675 }, { "epoch": 3.99, "learning_rate": 1.0417501402355959e-07, "loss": 0.0006, "step": 27680 }, { "epoch": 3.99, "learning_rate": 9.415818575206347e-08, "loss": 0.0006, "step": 27685 }, { "epoch": 3.99, "learning_rate": 8.414135748056736e-08, "loss": 0.001, "step": 27690 }, { "epoch": 3.99, "learning_rate": 7.412452920907125e-08, "loss": 0.001, "step": 27695 }, { "epoch": 4.0, "learning_rate": 6.410770093757512e-08, "loss": 0.0009, "step": 27700 }, { "epoch": 4.0, "learning_rate": 5.409087266607901e-08, "loss": 0.0008, "step": 27705 }, { "epoch": 4.0, "learning_rate": 4.40740443945829e-08, "loss": 0.0006, "step": 27710 }, { "epoch": 4.0, "learning_rate": 3.405721612308679e-08, "loss": 0.0011, "step": 27715 }, { "epoch": 4.0, "learning_rate": 2.4040387851590676e-08, "loss": 0.0007, "step": 27720 }, { "epoch": 4.0, "learning_rate": 1.402355958009456e-08, "loss": 0.0006, "step": 27725 }, { "epoch": 4.0, "learning_rate": 4.006731308598445e-09, "loss": 0.0007, "step": 27730 }, { "epoch": 4.0, "step": 27732, "total_flos": 6.633339766473071e+19, "train_loss": 0.0060005393999575125, "train_runtime": 13854.802, "train_samples_per_second": 128.116, "train_steps_per_second": 2.002 } ], "max_steps": 27732, "num_train_epochs": 4, "total_flos": 6.633339766473071e+19, "trial_name": null, "trial_params": null }